<?xml version="1.0" encoding="UTF-8"?>
<rfc xmlns:xi="http://www.w3.org/2001/XInclude" category="info" docName="draft-das-cross-platform-finality-authority-00" ipr="trust200902" submissionType="independent" xml:lang="en" version="3" tocInclude="true" tocDepth="3" sortRefs="true" symRefs="true">
<front>
<title abbrev="Interoperable Execution Finality">Interoperable Execution-Finality Authority Across Heterogeneous Computing Platforms: Protected Effect Semantics, Cross-Device Authority Continuity, and Resource-Side Enforcement</title>
<seriesInfo name="Internet-Draft" value="draft-das-cross-platform-finality-authority-00"/>
<author fullname="Sangam Kumar Das" initials="S." surname="Das">
<organization>Independent Inventor</organization>
<address>
<postal><city>Balasore</city><region>Odisha</region><code>756001</code><country>India</country></postal>
<phone>+91 98613 63532</phone>
<email>info@sangamdas.com</email>
</address>
</author>
<date year="2026" month="November" day="16"/>
<area>Security</area>
<workgroup>Independent Submission</workgroup>
<keyword>execution finality</keyword><keyword>cross-platform authorization</keyword><keyword>AI safety</keyword><keyword>agent security</keyword><keyword>Finality Sink</keyword><keyword>canonical effect descriptor</keyword>
<abstract>
<t>Computation is not authority to act. An operating system may authenticate an application, a device may recognize a permission, a cloud service may accept a token, an AI agent may hold a tool credential and a remote application may receive a command, yet none of these facts shows that the consequential operation actually executed is still within the originating authority, scope, current policy and resource state. This document describes an informational architecture for preserving an equivalent protected authorization relationship across heterogeneous platforms such as Android, iOS, macOS, Windows, Linux, cloud services, trusted hardware, edge devices and autonomous agents, whose native permission models and enforcement mechanisms differ. Qualified native evidence is translated into a protected canonical effect descriptor, scoped non-bearer enablement is created, and a receiving Finality Sink reconstructs the actual pending operation and enforces authority at a covered boundary. The document covers enforcement-capability qualification, cross-device ownership with source fencing, receipt-gated continuation, protected reauthorization, recovery and an honest account of what each mechanism can and cannot control.</t>
</abstract><note title="Source, Patent Basis and Notices">
<t>Foundational reference: PCT/IB2026/055615 ("Mothership"), filed 4 June 2026 and published 16 July 2026 as WO 2026/150382 A2. Where to read the full PCT filing: search "Sangam Das" on WIPO PATENTSCOPE; search PCT/IB2026/055615 (pctib202655615); or read the PCT file directly in Technical Disclosure Commons at https://www.tdcommons.org/dpubs_series/12542/.</t>
<t>Technical source of this document: an Indian provisional specification titled "Systems and Methods for Cross-Platform Interoperable Execution-Finality Control, Protected Authority Continuity, and Consequence Enforcement Across Heterogeneous Operating Systems, Devices, and Autonomous Computing Environments", supplied as a 3,577-page document with 40 distinct numbered embodiments and 60 claims, electronically signed 25 October 2026. The applicant identifies the application as 202631127058, filed 25 October 2026. The PDF text is not a certified Patent Office acknowledgment, so the number and date are applicant-identified and are not independently verified. See [[ipr]] and [[app-j]].</t>
<t>The IPR disclosure for this document will be made at the appropriate time in accordance with IETF guidelines.</t>
<t>Human-Originated Invention: the underlying invention is attributed to the human inventor Sangam Kumar Das. AI-Assisted Technical Preparation: this text was prepared with AI assistance and may contain errors; identifiers, dates, standards references and product statements must be verified against primary sources. Non-Endorsement: this is an individual proposal, not an IETF standard; company and product references are prospective research environments only, and no adoption, endorsement, partnership, copying, infringement or implementation is asserted. Security Limitation: protection applies only to effect paths actually covered; no universal operating-system privilege, instantaneous cross-domain revocation, global exactly-once execution, complete physical safety or zero-latency validation is presumed.</t>
</note></front>
<middle>
<section anchor="intro">
<name>Introduction</name>
<t>Interoperability between computing platforms is usually described as the ability to exchange data or requests. A different and harder question is whether two or more environments can preserve an equivalent protected authorization relationship for a consequential operation when their permission models, application isolation, hardware security, execution interfaces and trust assumptions differ.</t>
<t>The architecture answers that question without assuming uniform kernels or identical security APIs. Heterogeneous platforms translate qualified authority into operation-specific effect semantics, and a participating resource boundary independently enforces those semantics.</t>
<t>This document is understandable without reading the patent. Appendices map the mechanisms to the source specification and to examiner terminology; they are non-normative.</t>
</section>
<section anchor="problem">
<name>Problem Statement</name>
<t>A native permission on one platform and a native permission on another with a similar name may not mean the same thing. A credential accepted by a cloud service says who is calling, not which operation is about to take effect. When an unfinished operation moves from a phone to a cloud service, copying its state does not tell anyone whether the old device is still able to act. When a tool is updated, its effect semantics may change while its name stays the same. These problems are about authority for actual effects, not about transport or authentication.</t>
<t>The central question is how to keep protected authority meaningful across platforms, and how to say plainly where it stops.</t>
</section>
<section anchor="scope">
<name>Scope and Non-Goals</name>
<ul>
<li>In scope: cross-platform effect semantics, adapters, canonical descriptors, enforcement-capability qualification, scoped non-bearer enablement, destination-side reconstruction, authority consumption, cross-device ownership and fencing, revocation and offline limits, three common workflows (CW-1, CW-2, CW-3), recovery, and 40 embodiment profiles.</li>
<li>Not in scope: a universal wire format; a claim of non-bypassability beyond covered paths; global exactly-once execution; instantaneous revocation; certification; performance numbers; conclusions on novelty, priority or validity.</li>
<li>General concepts of Candidate Acts, Non-Effective State, Protected Enforcement Domains, Finality Sinks, scoped effect authority, revocation, anti-replay and actual-effect evidence are earlier work of the same programme and are not presented as first invented by the 25 October filing (<xref target="REL-FOUND"/>, <xref target="REL-AGENT"/>, <xref target="REL-IEV"/>, <xref target="REL-CBEA"/>, <xref target="PCT-055615"/>).</li>
</ul>
</section>
<section anchor="terms">
<name>Terminology</name>
<t>BCP 14 <xref target="RFC2119"/> <xref target="RFC8174"/> terms are used only where a requirement is intentionally stated.</t>
<dl newline="true">
<dt>Candidate Act</dt>
<dd><t>A proposed consequential operation held in a Non-Effective State.</t></dd>
<dt>Canonical Effect Descriptor</dt>
<dd><t>A versioned, typed representation of the material actual operation.</t></dd>
<dt>Finality Sink</dt>
<dd><t>The component at an actual resource boundary that reconstructs the pending operation and permits or withholds the effect.</t></dd>
<dt>Scoped non-bearer finality enablement</dt>
<dd><t>Operation-specific, sink-bound permission whose possession alone confers no authority.</t></dd>
<dt>Enforcement Qualification Record</dt>
<dd><t>A record of what a platform security mechanism can actually observe, block or consume.</t></dd>
<dt>Authority epoch</dt>
<dd><t>A monotonic generation used for ownership, revocation or policy.</t></dd>
<dt>Source fencing</dt>
<dd><t>Evidence or mechanism that prevents a previous owner from effecting the operation.</t></dd>
</dl>
</section>
<section anchor="trust">
<name>Cross-Platform Trust Model</name>
<t>The foundational sequence is: originating authority, Candidate Act, Non-Effective State, heterogeneous evidence translation, protected canonical effect descriptor, independent protected validation, scoped non-bearer finality enablement, receiving Finality Sink reconstruction, current authority and coverage check, protected consumption, actual consequence, and qualified outcome evidence. See <xref target="fig-whole"/>.</t>
<dl newline="true">
<dt>Source operating system</dt>
<dd><t>Controls the request origin; can prove what it observes, not what a remote sink will do.</t></dd>
<dt>Adapter</dt>
<dd><t>Translates native evidence; must itself be qualified; can mistranslate.</t></dd>
<dt>Protected Enforcement Domain</dt>
<dd><t>Holds authority, policy, revocation and consumption state; trusted only within its stated environment.</t></dd>
<dt>Finality Sink</dt>
<dd><t>Controls the effect at its boundary; cannot control paths it does not cover.</t></dd>
<dt>Outcome observer</dt>
<dd><t>Provides evidence within its observation capability.</t></dd>
</dl>
<t>An unauthorized action can still bypass the architecture through any effect path that does not pass a covered sink, through a compromised privileged host, or through a credential copy outside the broker. The draft therefore ties every assurance to the component that can provide it.</t>
</section>
<section anchor="semantics">
<name>Heterogeneous Authorization Semantics</name>
<t>Platform permission names are not treated as semantically equivalent. An adapter produces an effect-specific claim and records whether the claim is equivalent to, narrower than, or incompatible with the canonical semantics (<xref target="fig-xlate"/>). A native entitlement is advice; it is not final authority for an operation.</t>
<t>Syntactic interoperability (the fields parse) differs from semantic equivalence (the operation means the same). The former never implies the latter.</t>
</section>
<section anchor="candidate">
<name>Candidate Acts and Non-Effective State</name>
<t>A Candidate Act identifies the proposed operation and binds its action class, resource, destination, parameters, source authority and effect constraints. The Non-Effective State keeps the consequence from taking effect while conditions remain unsatisfied. Forms include protected queues, withheld credentials, uncommitted transactions and isolated command buffers. A simulated hold is insufficient if another effective route remains available. Where mandatory enforcement is unavailable, the defined condition is non-effective, not silently weakened.</t>
</section>
<section anchor="source">
<name>Source Authority and Delegation</name>
<t>The original authority, principal, delegated scope and restrictions are recorded. Delegation across providers keeps a cryptographic lineage in which scope can only narrow (<xref target="fig-deleg"/>). A forged or expanded delegation fails lineage verification. Source authentication is not the same as actual effect authority.</t>
<t>Embodiments E7, E28 and E29 address delegation continuity, transitive attenuation and approval quorums.</t>
</section>
<section anchor="adapters">
<name>Platform-Specific Evidence Adapters</name>
<t>Adapters translate qualified native-platform security information into the protected common representation. For each platform below, the table-like list states what established controls provide, what additional binding is needed, and what is not assumed.</t>
<section>
<name>Android</name>
<dl newline="true">
<dt>Established native controls</dt>
<dd><t>Application UID isolation, app permissions, Keystore and hardware-backed keys, StrongBox where available, managed/enterprise device services, app-controlled tool dispatchers, server-side resource enforcement, authorised accessibility interfaces, device state and security updates.</t></dd>
<dt>Additional binding needed for effect-specific cross-platform authority</dt>
<dd><t>Native permissions describe what an app may ask the platform to do; they do not describe the specific consequential operation. A binding of permission, operation-specific canonical descriptor and a resource-side check is needed.</t></dd>
<dt>Not assumed</dt>
<dd><t>An ordinary Android application has no unrestricted system-wide mediation; enforcement is limited to effects routed through the app dispatcher or a participating resource-side gateway.</t></dd>
</dl>
</section>
<section>
<name>iOS, iPadOS and macOS</name>
<dl newline="true">
<dt>Established native controls</dt>
<dd><t>App sandbox, entitlements, Keychain, Secure Enclave, app-controlled effectuation, system-approved extensions, managed apps or enterprise interfaces, server-side enforcement, app-to-app and device-to-device communication, current platform security state.</t></dd>
<dt>Additional binding needed for effect-specific cross-platform authority</dt>
<dd><t>The same operation-specific binding is needed; platform identity of the app is not authority for a particular effect.</t></dd>
<dt>Not assumed</dt>
<dd><t>No private kernel hooks, no universal app interception and no access to other apps credentials are assumed.</t></dd>
</dl>
</section>
<section>
<name>Windows</name>
<dl newline="true">
<dt>Established native controls</dt>
<dd><t>Process and service identities, AppContainer and isolation mechanisms, security principals, credential protection, TPM-backed keys, protected service brokers, endpoint management, privileged execution interfaces.</t></dd>
<dt>Additional binding needed for effect-specific cross-platform authority</dt>
<dd><t>A broker or service can hold authority, but binding it to the exact pending effect and to a revocation generation is an addition.</t></dd>
<dt>Not assumed</dt>
<dd><t>Security principals do not by themselves limit what an authorised principal does with a credential it can export.</t></dd>
</dl>
</section>
<section>
<name>Linux</name>
<dl newline="true">
<dt>Established native controls</dt>
<dd><t>Process credentials, namespaces, LSMs, seccomp, capabilities, cgroups, containers, hardware-backed keys, service-mesh and workload identity, privileged resource controllers.</t></dd>
<dt>Additional binding needed for effect-specific cross-platform authority</dt>
<dd><t>Kernel controls restrict processes; they do not know an operation-level canonical effect. A resource controller or sink must reconstruct the operation.</t></dd>
<dt>Not assumed</dt>
<dd><t>Application-level code cannot independently resist a fully compromised kernel.</t></dd>
</dl>
</section>
<section>
<name>Cloud and embedded</name>
<dl newline="true">
<dt>Established native controls</dt>
<dd><t>Cloud IAM, resource-native API authorization, managed identity, confidential computing, HSM/KMS, secure and measured boot, device command controllers, firmware security, trusted peripherals, remote attestation, resource-gateway enforcement.</t></dd>
<dt>Additional binding needed for effect-specific cross-platform authority</dt>
<dd><t>Cloud IAM authorizes API principals; the additional binding is operation-specific, sink-reconstructed and consumed once under protected state.</t></dd>
<dt>Not assumed</dt>
<dd><t>Where a direct credential path to the resource remains, a gateway does not control the effect.</t></dd>
</dl>
</section>
</section>
<section anchor="canon">
<name>Canonical Effect Descriptors</name>
<t>The Canonical Effect Descriptor is versioned and typed. Candidate fields: descriptor version; original operation and Candidate Act identifiers; originating principal; requesting workload; source and receiving platform; parent task; delegation chain; effect class; effective resource identity; destination or beneficiary; material parameters; allowed transformations; source authority; scope; policy, revocation and resource generations; designated Finality Sink; required evidence; enforcement-coverage reference; validity conditions; protected usage conditions. Mandatory fields depend on the effect; not every application carries all fields.</t>
<t>Canonicalization must address Unicode normalization, binary addresses, object ordering, numeric precision, timestamps, resource aliases, URL normalization, redirects, nested operations, unknown critical fields, schema evolution, units, localized strings and application-specific semantics. Unknown critical fields cause rejection. A valid hash of an incorrectly built descriptor does not prove the actual operation has the intended meaning (<xref target="fig-desc"/>).</t>
<t>Illustrative JSON (not a mandatory encoding):</t>
<figure anchor="ex-desc">
<name>Illustrative canonical effect descriptor</name>
<sourcecode type="json"><![CDATA[{
  "v": 1,
  "op": "op-7f3a",
  "candidate_act": "ca-19",
  "effect_class": "payment.transfer",
  "resource": "acct:EXAMPLE-1",
  "beneficiary": "acct:EXAMPLE-2",
  "params": {"amount": "100.00", "ccy": "EUR"},
  "source_platform": "android",
  "recv_platform": "linux",
  "authority_gen": 12,
  "revocation_gen": 4,
  "sink": "sink-pay-01",
  "critical": ["resource","beneficiary","params"]
}]]></sourcecode>
</figure>
</section>
<section anchor="qualification">
<name>Enforcement-Capability Qualification</name>
<t>A security mechanism may provide authentication, authorization advice, logging, observation, notification, isolation, revocation notification, restricted credential use, operation-specific blocking, actual resource-side refusal, protected authority consumption or trusted state verification. These capabilities are not interchangeable. A monitoring API is not a blocking interface because its name contains security; a sandbox does not control an external service that accepts the agent's credentials; a credential broker does not control other copies of credentials (<xref target="fig-qual"/>, <xref target="fig-obs"/>).</t>
<t>An Enforcement Qualification Record contains: platform; security component; controlled resource; operation class; available observation; ability to block; current component generation; required trust assumption; known alternate effect routes; qualification status; and the event that requires revalidation. Status values are qualified, partially qualified, unresolved and disqualified. Where mandatory blocking is unavailable, the options are to move enforcement to a participating resource-side gateway, to restrict supported effects, or to withhold the protected operation. The claimed protection property is not silently weakened.</t>
</section>
<section anchor="ped">
<name>Protected Enforcement Domains</name>
<t>The Protected Enforcement Domain maintains applicable authority, evidence, policy versions, revocation, consumption counters and technical conditions. An Extended Verification Module Interface supports optional checks on semantic identity, privacy, conflicts, device state, resource limits, evidence, delegation or prior-effect dependencies. A Protected Authority Controller creates, derives or activates scoped finality enablement; a Finality Gate combines mandatory predicates. Domain separation is by control, not by location alone.</t>
</section>
<section anchor="enablement">
<name>Scoped Finality Enablement</name>
<t>Enablement is operation-specific, sink-bound and non-bearer: possessing it does not authorize arbitrary use (<xref target="fig-nonbearer"/>). It is bound to the descriptor, the designated sink, current generations and a protected usage condition. A sink that cannot verify the binding refuses.</t>
<t>Local authority and the destination sink are distinct (<xref target="fig-local"/>). A credential broker can hold the credential, but downstream enforcement must still check the operation (<xref target="fig-broker"/>).</t>
</section>
<section anchor="sinks">
<name>Receiving-Platform Finality Sinks</name>
<t>The sink independently obtains or reconstructs the actual pending operation, checks its correspondence to the descriptor and to current sink-local state, and permits or withholds the effect through a controlled enforcement boundary. A sink that only logs is not a sink. A sink controls only effects that pass through it; effect-path discovery and a coverage controller report which routes are covered.</t>
</section>
<section anchor="consumption">
<name>Authority Consumption</name>
<t>Authorization is consumed under protected monotonic state (<xref target="fig-replay"/>). A crash between consumption and effect leaves the outcome indeterminate; it is recorded as unknown and is not blindly retried. Authority and revocation epochs are checked at effect time (<xref target="fig-epoch"/>).</t>
</section>
<section anchor="migration">
<name>Cross-Device Ownership and Migration</name>
<t>The stages are: identify the unfinished operation; establish the current source owner; authenticate the target; verify eligibility; determine effect-capable source routes; prepare transferable computational context; distinguish computational state from finality authority; establish a protected ownership-transition generation; verify source fencing; reconstruct the destination effect scope; establish destination-specific protected authority; verify the actual pending operation at the destination; prevent stale source effectuation; preserve consumed or unresolved effect state; reconcile partial migration failure (<xref target="fig-migr"/>, <xref target="fig-fence"/>).</t>
<t>Negative scenarios include a suspended or disconnected source, both sides active, snapshot restoration reviving an old worker, a source that retains an API credential, a hardware queue that remains executable, weaker destination trust, an effect before transfer completes, revocation during transfer, a delayed callback to the source, destination retry of an unresolved operation, and accidental merging of two pending operations.</t>
<t>Invariant: no duplicated, unauthorized execution authority may arise merely from copying or restoring computational state. Instant global revocation across disconnected devices is not assumed.</t>
</section>
<section anchor="revocation">
<name>Revocation and Offline Operation</name>
<t>Revocation propagates across domains and invalidates dependent authority; offline authority is bounded by explicit leases and clock uncertainty (<xref target="fig-offline"/>). Cached permission after revocation, epoch rollback and incorrect lease expiry are threats listed in <xref target="app-e"/>. Delayed revocation is a stated limitation, not a solved problem.</t>
</section>
<section anchor="cw1">
<name>Common Workflow CW-1</name>
<t>CW-1 is the modular protected architecture with actual-operation reconstruction, independent protected authority, non-bypassable consequence enforcement and optional advanced mechanisms. Components: Requesting Computational Entity (constructs requests, holds no finality authority automatically); Candidate Act Constructor; Non-Effective Holding Mechanism; Protected Enforcement Domain; Extended Verification Module Interface; Protected Authority Controller; Finality Gate; Finality Sink; Protected State and Evidence Store; Optional Effect Observer; Optional Interim Effectuation Validator; Effectuation-Path Coverage Controller. No implementation need contain every optional module.</t>
</section>
<section anchor="cw2">
<name>Receipt-Gated Workflow CW-2</name>
<t>CW-2 divides an operation into bounded phases. The first phase is independently authorized; a bounded real effect occurs through a participating sink; qualified evidence is obtained; an authorized Interim Effectuation Validator evaluates it; the accepted receipt becomes a required causal condition for the next phase; the successor sink independently checks the receipt-dependent authority. Missing, stale, replayed, unqualified or contradicted evidence does not enable the successor effect (<xref target="fig-cw2a"/>, <xref target="fig-cw2b"/>).</t>
<ul>
<li>A test simulation is not a real first effect.</li>
<li>A tool-generated success message is not automatically qualified evidence.</li>
<li>A signed receipt is not automatically true outside the signer's observation capability.</li>
<li>A validator need not relay the original command and is not an unrestricted credential.</li>
<li>Receipts bind to phase, sink, operation, state and authority context.</li>
<li>CW-2 is optional where continuation does not need a receipt.</li>
</ul>
</section>
<section anchor="cw3">
<name>Reauthorization Workflow CW-3</name>
<t>CW-3 covers original task scope, cumulative actual effect, material scope deviation, independent security signals, false positives, conflicting evidence, risk-adaptive suspension, reauthorization eligibility, the exact changed consequence, human or authorized automated approval, protected authority generation, current policy and revocation, limited renewed authority, sink-side verification, repeated deviations, recovery and audit (<xref target="fig-cw3"/>).</t>
<t>A human approval or automated reauthorization cannot override unrelated mandatory protections: approval to continue a robotic operation does not establish physical safety, actuator identity or valid prior-effect evidence. An approval screen, a confirmation prompt or an alert acknowledgment is insufficient unless the resulting permission is bound to the actual effect.</t>
</section>
<section anchor="impl">
<name>Present-Day Implementability (CW-IMP-1)</name>
<t>The source's supplementary implementation material (CW-IMP-1) explains how existing technologies can support selected functions: protected application dispatchers, credential brokers, secure enclaves, TEEs, HSMs, secure elements, transactional databases, authenticated receipts, hardware command gates, server-side resource controllers, attestation services, runtime policy engines, monotonic state, recovery journals, isolated services and protected device communication. These provide building blocks; the additional enforcement relationships (operation reconstruction at the sink, operation-bound enablement, source fencing) are what the architecture adds. For every assurance, the component able to provide it is named in <xref target="app-d"/> and in the mechanism list in <xref target="hw"/>. Application code cannot resist a fully compromised kernel; an iOS app has no arbitrary system-wide interception; an Android app cannot mediate all device or app operations; cross-platform non-bypassability is not claimed unless all relevant paths are controlled.</t>
</section>
<section anchor="tool">
<name>Tool Runtime and Agent Handoff</name>
<t>Tool installation, update and runtime changes are consequential transitions: authority bound to a prior generation does not carry over without revalidation (<xref target="fig-toolup"/>). Tool outputs are data, not instruction authority. Cross-provider handoff keeps delegation lineage; moving agent context does not move effect authority. Nested tool invocation attenuates authority. Embodiments E10, E20, E24, E25, E26, E27, E28 and E34 apply.</t>
<t>Tool names that look equivalent may carry different permissions or effects; compatibility is validated per contract version before authority is accepted.</t>
</section>
<section anchor="hw">
<name>Hardware and Cloud Enforcement</name>
<t>For each enforcement alternative the entries state: effects it can control; what an attacker can bypass; whether protected state survives crashes; whether it can prevent replay; whether it can verify current revocation; whether authority is transferable; whether it can attest its enforcement generation; and what happens when it is unavailable. Not every implementation is hardware-rooted, and software-only separation from a fully compromised privileged host is not claimed.</t>
<section>
<name>Application-level enforcement</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Only effects routed through the application dispatcher.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>A compromised or modified application, or any direct credential path.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Only if the application journals durably.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Only with a durable counter.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Only by querying a service.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Only by explicit transfer protocol.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>No, unless backed by an attestable component.</t></dd>
<dt>If unavailable</dt>
<dd><t>Effects through the dispatcher stop; other paths are unaffected.</t></dd>
</dl>
</section>
<section>
<name>Privileged OS service</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Effects mediated by the service on that host.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>A compromised kernel or direct driver access.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Depends on service storage.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Yes, if state is monotonic.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>If it can reach the revocation source.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Not inherently.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Possible with platform attestation.</t></dd>
<dt>If unavailable</dt>
<dd><t>Mediated effects stop; the rest is outside control.</t></dd>
</dl>
</section>
<section>
<name>Managed device gateway</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Effects of managed devices routed through it.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Unmanaged devices and alternate networks.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Yes if stored in the gateway.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Yes.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Yes when online.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Gateway-held authority can be moved.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Partly.</t></dd>
<dt>If unavailable</dt>
<dd><t>Managed effects stop; others unaffected.</t></dd>
</dl>
</section>
<section>
<name>Remote resource controller</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Effects at the resource it fronts.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Paths to the resource that bypass the controller.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Yes if transactional.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Yes.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Yes when it can reach revocation state.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Resource-bound, not transferable.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Possible.</t></dd>
<dt>If unavailable</dt>
<dd><t>The resource refuses by default; availability falls.</t></dd>
</dl>
</section>
<section>
<name>Hardware security module</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Cryptographic operations it holds keys for.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Effects outside the key use.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Yes, in the module.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Via monotonic counters.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Not by itself.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Keys may be non-transferable.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Yes.</t></dd>
<dt>If unavailable</dt>
<dd><t>Key use stops.</t></dd>
</dl>
</section>
<section>
<name>Secure element</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Operations the element signs or authorizes.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Everything not routed through it.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Yes.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Via counters.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Not by itself.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Typically bound to a device.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Yes.</t></dd>
<dt>If unavailable</dt>
<dd><t>Key use stops.</t></dd>
</dl>
</section>
<section>
<name>TPM-backed authority</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Operations needing keys sealed to platform state.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Effects not gated by those keys.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Yes for sealed state.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Via counters.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>No.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Bound to platform.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Yes (measured state).</t></dd>
<dt>If unavailable</dt>
<dd><t>Sealed secrets become unusable.</t></dd>
</dl>
</section>
<section>
<name>Confidential computing</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Computation inside the protected environment.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Effects emitted after leaving it.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Depends on storage design.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Possible.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Not by itself.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Depends.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Yes, for the environment.</t></dd>
<dt>If unavailable</dt>
<dd><t>Environment attestation fails closed.</t></dd>
</dl>
</section>
<section>
<name>Trusted execution environment</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Operations run in the trusted world.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>The normal world and unrelated paths.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Depends.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Possible.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Not by itself.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Depends.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Yes.</t></dd>
<dt>If unavailable</dt>
<dd><t>Trusted operations stop.</t></dd>
</dl>
</section>
<section>
<name>Protected virtual machine</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Effects the VM boundary mediates.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Host-level compromise and paths outside the VM.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Depends.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Possible.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Not by itself.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Migration needs fencing.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Depends.</t></dd>
<dt>If unavailable</dt>
<dd><t>VM-mediated effects stop.</t></dd>
</dl>
</section>
<section>
<name>Hardware command buffer</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Queued device commands.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Direct register or alternate queue access.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Depends on device.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Depends.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>No.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Not transferable.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Device-dependent.</t></dd>
<dt>If unavailable</dt>
<dd><t>Queued commands are withheld.</t></dd>
</dl>
</section>
<section>
<name>GPU or accelerator command control</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Commands through the controlled queue.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Alternate queues or drivers.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Depends.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Depends.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>No.</t></dd>
<dt>Authority transferable</dt>
<dd><t>No.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Device-dependent.</t></dd>
<dt>If unavailable</dt>
<dd><t>Controlled commands stop.</t></dd>
</dl>
</section>
<section>
<name>Network egress controller</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Traffic crossing the egress point.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Alternate egress paths.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Depends.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Yes for counters.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>Yes if connected.</t></dd>
<dt>Authority transferable</dt>
<dd><t>Not transferable.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Possible.</t></dd>
<dt>If unavailable</dt>
<dd><t>Egress blocks; connectivity falls.</t></dd>
</dl>
</section>
<section>
<name>Secure display and user confirmation</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>What the user sees and confirms.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Differences between displayed and executed operation.</t></dd>
<dt>State survives crashes</dt>
<dd><t>N/A.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Single-use confirmation tokens.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>N/A.</t></dd>
<dt>Authority transferable</dt>
<dd><t>No.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Platform dependent.</t></dd>
<dt>If unavailable</dt>
<dd><t>No confirmation, no operation.</t></dd>
</dl>
</section>
<section>
<name>Physical actuator controller</name>
<dl newline="true">
<dt>Effects it can control</dt>
<dd><t>Commands to the actuator.</t></dd>
<dt>Attacker bypass</dt>
<dd><t>Direct wiring or an uncontrolled controller.</t></dd>
<dt>State survives crashes</dt>
<dd><t>Yes if non-volatile.</t></dd>
<dt>Prevents replay</dt>
<dd><t>Yes.</t></dd>
<dt>Verifies current revocation</dt>
<dd><t>No.</t></dd>
<dt>Authority transferable</dt>
<dd><t>No.</t></dd>
<dt>Attests its enforcement generation</dt>
<dd><t>Possible.</t></dd>
<dt>If unavailable</dt>
<dd><t>Actuation withheld; safety state must be defined separately.</t></dd>
</dl>
</section>
</section>
<section anchor="failure">
<name>Failure and Recovery</name>
<t>Unknown completion stays unknown. Replay of consumed authority is refused. Recovery is a new Candidate Act with its own authority; compensation is separately authorized. Partial multi-resource effects are contained and reconciled without promising atomicity (<xref target="fig-multi"/>, <xref target="fig-recov"/>). Irreversible effects are checked for hazard before effect and, if unsafe, contained (<xref target="fig-irrev"/>). Evidence from a possibly compromised observer is checked by independent attestation (<xref target="fig-obsrv"/>).</t>
</section>
<section anchor="clusters">
<name>Embodiment Clusters</name>
<t>Six clusters group the embodiments. Embodiment numbering E1 to E40 follows the source; the second full section labelled Embodiment 31 is a repeated version, not a forty-first embodiment (<xref target="app-j"/>).</t>
<section anchor="cl-a">
<name>Cluster A: Effect Conflict, Quorums and Cumulative Resources</name>
<t>Embodiments: E1, E5, E11, E13, E22, E29.</t>
<t>A quorum approval is not automatically authority to execute. A resource fairness decision is not automatically a valid execution capability. Individually permissible operations may become impermissible when jointly applied; examples are cloud resource allocation, network reconfiguration and payment budgets.</t>
<ul>
<li>E1: Protected cross-agent effect-conflict arbitration with atomic reservations and sink-bound fencing</li>
<li>E5: Cryptographically verifiable, assistant-neutral resource allocation and protected eligible-demand completeness</li>
<li>E11: Concurrent-effect reconciliation, version-bound reservations and sink-local commit validation</li>
<li>E13: Multi-agent consensus, quorum integrity and non-delegable finality authorization</li>
<li>E22: Multi-resource budgets, delegated quotas and cumulative consequential limits</li>
<li>E29: Independent approval quorums and protected consensus-to-effect separation</li>
</ul>
</section>
<section anchor="cl-b">
<name>Cluster B: Receipts, Evidence and Outcome Truth</name>
<t>Embodiments: E2, E4, E15, E16, E38, E40.</t>
<t>A signature is not truth, and a dispatch acknowledgment is not actual completion. Independent validator identity, observation coverage, consumed authorization and indeterminate outcomes must each be represented.</t>
<ul>
<li>E2: Identity-bound Interim Effectuation Validation and receipt-constituted causal continuation</li>
<li>E4: Cross-domain actual-consequence witnessing and qualified remote-effect evidence</li>
<li>E15: Cross-domain evidence commitments, non-bearer authority and sink-local monotonic consumption</li>
<li>E16: Crash-consistent finality journaling, one-time authorization and indeterminate outcome handling</li>
<li>E38: Evidence integrity, compromised observers and independent attestation reconstruction</li>
<li>E40: End-to-end effect provenance, protected authority consumption, cross-embodiment closure and formal verification</li>
</ul>
</section>
<section anchor="cl-c">
<name>Cluster C: Semantic, Privacy and Resource Identity</name>
<t>Embodiments: E3, E6, E8, E19, E21, E24, E32, E34.</t>
<t>The visible object, the named tool and the identifier may not identify the actual effective resource after rebinding, redirect or contract change. Human approval is meaningful only when bound to the actual operation.</t>
<ul>
<li>E3: User-referenced semantic object correspondence and authenticated presentation-state binding</li>
<li>E6: Semantic action-contract evolution and effect-compatibility validation</li>
<li>E8: Correlation-aware cumulative privacy exposure and compositional disclosure constraints</li>
<li>E19: Computer-use and transaction-bound human-approval finality</li>
<li>E21: Actual resource identity, aliases, redirects and dynamic namespace rebinding</li>
<li>E24: Tool-output and instruction provenance through model-driven execution</li>
<li>E32: Data, memory and persistent-knowledge provenance across agent sessions</li>
<li>E34: Tool and resource identity continuity across dynamic service discovery</li>
</ul>
</section>
<section anchor="cl-d">
<name>Cluster D: Authority Across Devices and Organizations</name>
<t>Embodiments: E7, E9, E17, E18, E27, E30, E31, E35, E39.</t>
<t>Moving computational state does not establish transfer of effect authority. Loss of source-device communication does not prove the source has been fenced. Offline validity and delayed revocation must be explicit.</t>
<ul>
<li>E7: Cross-provider delegated agent authority continuity and cryptographic delegation lineage</li>
<li>E9: Cross-device execution-authority ownership, hardware epochs and source fencing</li>
<li>E17: Temporal authority, lease expiration, clock uncertainty and delayed-effect control</li>
<li>E18: Cross-domain revocation propagation and dependent authority invalidation</li>
<li>E27: Distributed agent handoff, workload migration and non-resurrectable in-flight effects</li>
<li>E30: Resource-owner changes, competing administrative authority and revocation races</li>
<li>E31: Credential and administrative authority rotation, protected key custody and generation fencing</li>
<li>E35: Degraded connectivity, offline effect authority and delayed revocation</li>
<li>E39: Conflicting policies and authority precedence among multiple principals and resource owners</li>
</ul>
</section>
<section anchor="cl-e">
<name>Cluster E: Tool Execution, Sandboxes and Runtime Changes</name>
<t>Embodiments: E10, E14, E20, E23, E25, E26, E28, E33.</t>
<t>Blocking one API route does not prevent an independent direct-credential route. Runtime or policy changes require mandatory revalidation, and a tool output is not an instruction authority.</t>
<ul>
<li>E10: Model, algorithm and runtime-generation drift with effectuation-time revalidation</li>
<li>E14: Effectuation-path discovery, bypass detection and enforcement-coverage revalidation</li>
<li>E20: Multi-step tool-chain and workflow dependency-bound execution authority</li>
<li>E23: Credential and secret-use finality with non-exportable broker-controlled authority</li>
<li>E25: Agent sandbox privilege transitions and consequential operations escaping sandbox boundaries</li>
<li>E26: Tool installation, updates and runtime lifecycle authority continuity</li>
<li>E28: Nested multi-agent tool invocation chains and transitive authority attenuation</li>
<li>E33: Policy and configuration evolution, dependency impact and non-self-authorizing mutations</li>
</ul>
</section>
<section anchor="cl-f">
<name>Cluster F: Recovery and Irreversible Effects</name>
<t>Embodiments: E12, E16, E18, E27, E35, E36, E37.</t>
<t>Unknown completion, replay and partial multi-resource effects must stay visible. Compensation is a separately authorized Candidate Act. No universal rollback or distributed atomicity is promised.</t>
<ul>
<li>E12: Autonomous recovery, compensating effects and non-replayable recovery authority</li>
<li>E16: Crash-consistent finality journaling, one-time authorization and indeterminate outcome handling</li>
<li>E18: Cross-domain revocation propagation and dependent authority invalidation</li>
<li>E27: Distributed agent handoff, workload migration and non-resurrectable in-flight effects</li>
<li>E35: Degraded connectivity, offline effect authority and delayed revocation</li>
<li>E36: Multi-resource transactions, partial commits and independently authorized compensation</li>
<li>E37: Irreversible effects, consequence-bound hazard checks and non-compensable operation containment</li>
</ul>
</section>
</section>
<section anchor="examples">
<name>Worked Examples</name>
<t>The 17 worked examples follow the chain: original authority, Candidate Act, platform-native inputs, canonical effect, protected verification, destination Finality Sink, outcome evidence, failure and recovery. They are illustrations, not test results.</t>
<section>
<name>Example A: Android app requests an operation on a protected iOS-associated service.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Android app permission; user intent</t></dd>
<dt>Candidate Act</dt>
<dd><t>Candidate Act held non-effective</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Android UID and Keystore evidence</t></dd>
<dt>Canonical effect</dt>
<dd><t>Canonical descriptor: service, resource, parameters</t></dd>
<dt>Protected verification</dt>
<dd><t>Descriptor validated; scoped enablement</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>iOS-side sink reconstructs request and consumes enablement</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Sink-side evidence of the effect</t></dd>
<dt>Failure and recovery</dt>
<dd><t>If the sink cannot reconstruct: refuse; unknown outcome stays UNRESOLVED</t></dd>
</dl>
</section>
<section>
<name>Example B: iOS task delegated to an Android device while authority stays bounded.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>iOS user authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Candidate Act at the delegate</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Delegation lineage</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with attenuated scope</t></dd>
<dt>Protected verification</dt>
<dd><t>Lineage verified</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Android sink checks bounded scope</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Receipt</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Scope expansion: refuse</t></dd>
</dl>
</section>
<section>
<name>Example C: Windows cloud-management app requests a Linux server configuration change.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Windows service identity</t></dd>
<dt>Candidate Act</dt>
<dd><t>Config change Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Principal and device evidence</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor: host, file, diff</t></dd>
<dt>Protected verification</dt>
<dd><t>Policy and revocation checked</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Linux resource controller reconstructs the change</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Controller evidence</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Controller unreachable: stay non-effective</t></dd>
</dl>
</section>
<section>
<name>Example D: Agent migrates from phone to cloud with an unresolved financial transaction.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Phone-held authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Pending Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Source state and fencing evidence</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor of the payment</t></dd>
<dt>Protected verification</dt>
<dd><t>New ownership generation</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Cloud sink verifies destination authority</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Outcome evidence for the first attempt</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Unknown outcome: do not retry blindly</t></dd>
</dl>
</section>
<section>
<name>Example E: Source device retains a credential after migration.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Source authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Pending act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Credential custody evidence</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor bound to generation</t></dd>
<dt>Protected verification</dt>
<dd><t>Generation fenced</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink rejects old generation</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Rejection event</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Credential exported: sink refuses it</t></dd>
</dl>
</section>
<section>
<name>Example F: Tool permissions differ between providers though names match.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Provider A tool contract</t></dd>
<dt>Candidate Act</dt>
<dd><t>Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Contract versions</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with contract identifier</t></dd>
<dt>Protected verification</dt>
<dd><t>Compatibility validation fails</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink refuses</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Refusal</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Incompatible: withhold</t></dd>
</dl>
</section>
<section>
<name>Example G: Model-generated instruction exports a sensitive file to a redirected destination.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Task authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Export Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Instruction provenance</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with effective destination</t></dd>
<dt>Protected verification</dt>
<dd><t>Provenance and destination check</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Egress sink checks actual destination</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Refusal</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Redirect: refuse</t></dd>
</dl>
</section>
<section>
<name>Example H: Resource alias changes between approval and effect.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Human approval</t></dd>
<dt>Candidate Act</dt>
<dd><t>Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Aliased referent at approval</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with resolved identity</t></dd>
<dt>Protected verification</dt>
<dd><t>Revalidate at effect time</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink resolves alias</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Evidence of resolution</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Changed referent: refuse</t></dd>
</dl>
</section>
<section>
<name>Example I: App changes request parameters after approval.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Approval</t></dd>
<dt>Candidate Act</dt>
<dd><t>Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Approved parameters</t></dd>
<dt>Canonical effect</dt>
<dd><t>Approved descriptor</t></dd>
<dt>Protected verification</dt>
<dd><t>Digest binds parameters</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink reconstructs parameters</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Mismatch evidence</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Different parameters: refuse</t></dd>
</dl>
</section>
<section>
<name>Example J: Signed receipt falsely claims a remote effect.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>First-phase authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Phase 1 act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Receipt</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor of phase 1</t></dd>
<dt>Protected verification</dt>
<dd><t>Validator checks signer capability</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Successor sink refuses continuation</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Unverified receipt recorded</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Receipt outside observation capability: not accepted</t></dd>
</dl>
</section>
<section>
<name>Example K: First bounded effect succeeded but independent receipt unavailable.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Phase 1 authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Phase 2 act held</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>No receipt</t></dd>
<dt>Canonical effect</dt>
<dd><t>Phase 2 descriptor</t></dd>
<dt>Protected verification</dt>
<dd><t>Receipt required</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Successor sink holds</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Pending record</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Remain non-effective until receipt or reauthorization</t></dd>
</dl>
</section>
<section>
<name>Example L: Human reauthorizes a changed action but a physical safety condition is unsatisfied.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Reauthorization</t></dd>
<dt>Candidate Act</dt>
<dd><t>Revised Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Approval and hazard evidence</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with hazard fields</t></dd>
<dt>Protected verification</dt>
<dd><t>Hazard check fails</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Actuator gate refuses</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Refusal</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Approval does not override safety</t></dd>
</dl>
</section>
<section>
<name>Example M: Credential broker protected while an alternate unrestricted key remains.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Broker authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Broker and key inventory</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor</t></dd>
<dt>Protected verification</dt>
<dd><t>Coverage controller flags alternate key</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink only controls brokered path</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Coverage report</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Claim limited; alternate path disclosed</t></dd>
</dl>
</section>
<section>
<name>Example N: Disconnected device exercises stale authority.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Cached authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Offline act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Offline limits and clock bounds</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with lease</t></dd>
<dt>Protected verification</dt>
<dd><t>Limits exceeded</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink refuses</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Refusal</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Offline past limit: refuse</t></dd>
</dl>
</section>
<section>
<name>Example O: Two agents collectively exceed a protected budget.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Agent grants</t></dd>
<dt>Candidate Act</dt>
<dd><t>Two acts</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Budget state</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptors with budget refs</t></dd>
<dt>Protected verification</dt>
<dd><t>Atomic reservation: second fails</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink applies reservation</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Reservation record</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Second request refused</t></dd>
</dl>
</section>
<section>
<name>Example P: Tool update changes effect semantics mid-workflow.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Workflow authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Next step act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Tool generation</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with generation</t></dd>
<dt>Protected verification</dt>
<dd><t>Mismatch triggers revalidation</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Sink compares generation</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Revalidation record</t></dd>
<dt>Failure and recovery</dt>
<dd><t>Stop at the step boundary</t></dd>
</dl>
</section>
<section>
<name>Example Q: Physical command is irreversible and cannot safely be compensated.</name>
<dl newline="true">
<dt>Original authority</dt>
<dd><t>Operator authority</t></dd>
<dt>Candidate Act</dt>
<dd><t>Candidate Act</t></dd>
<dt>Platform-native inputs</dt>
<dd><t>Hazard and state evidence</t></dd>
<dt>Canonical effect</dt>
<dd><t>Descriptor with irreversibility flag</t></dd>
<dt>Protected verification</dt>
<dd><t>Pre-effect hazard check passes or fails</t></dd>
<dt>Destination Finality Sink</dt>
<dd><t>Actuator controller gate</t></dd>
<dt>Outcome evidence</dt>
<dd><t>Effect observation</t></dd>
<dt>Failure and recovery</dt>
<dd><t>If unsafe: contain; no compensation promised</t></dd>
</dl>
</section>
</section>
<section anchor="interop">
<name>Interoperability Requirements</name>
<t>Where an implementation claims interoperability under this profile, it should carry the fields marked mandatory for the effect class, reject unknown critical fields, use qualified adapters, record qualification status, and treat unavailable enforcement as non-effective. It must not claim protection for effect paths that are not covered. No universal encoding is mandated; CBOR <xref target="RFC8949"/> and JSON are illustrative choices.</t>
<t>Records are listed in <xref target="app-d"/>.</t>
</section>
<section anchor="security">
<name>Security Considerations</name>
<t>Guidance follows <xref target="RFC3552"/>. The architecture does not protect against effects through uncovered paths, a compromised privileged host when enforcement runs on that host, or false evidence from an unqualified source. A cryptographic signature or zero-trust label does not by itself prevent any listed threat; each threat in <xref target="app-e"/> names the check that is required. Enforcement qualification records themselves are an attack surface and need freshness and integrity. Resource exhaustion by forcing unresolved states is possible and needs bounded handling.</t>
</section>
<section anchor="privacy">
<name>Privacy Considerations</name>
<t>Considerations in <xref target="RFC6973"/> apply. Separate valid disclosures may collectively create an unauthorized aggregate disclosure under a policy, so cumulative exposure budgets are tracked (E8). Agent memory, context compression and persistent knowledge carry provenance (E32). A privacy-preserving log does not by itself stop a release; disclosure control must identify the real egress or destination boundary. Execution evidence can itself leak information, so audit minimization and selective disclosure are needed.</t>
</section>
<section anchor="perf">
<name>Performance and Deployment</name>
<t>No performance results are claimed. Three paths are distinguished: a fast path (local protected finality checks and current protected-state admission), a warm path (independent policy or semantic validation, resource identity resolution, authority verification, evidence qualification) and a recovery path (reconciliation, revocation propagation, migration, unresolved outcomes, reauthorization). There is no per-token validation, no model-inference interruption, and no remote validator call for harmless operations; enforcement sits at real consequential boundaries. Protected precomputation and local gates can reduce latency. Cross-domain consensus has inherent cost, and behaviour during disconnection must be explicit.</t>
<t>A validation plan should measure p50, p95, p99 and p99.9 latency, throughput, memory, energy and battery impact, validator availability, protected-state persistence and recovery correctness on the target hardware.</t>
</section>
<section anchor="prior">
<name>Prior Art and Existing Standards</name>
<t>For each mechanism: it already provides authentication, binding or attestation; the remaining problem is keeping authority meaningful for an actual effect across heterogeneous platforms; the additional mechanisms described are operation-specific descriptors, qualification records, sink reconstruction, consumption and fencing; whether the literature already provides substantially the same relationship was not checked here; evidence for any improvement would need testing.</t>
<table>
<thead><tr><th>Established mechanism</th><th>What it provides and what remains</th></tr></thead>
<tbody>
<tr><td>OAuth <xref target="RFC6749"/>, token exchange <xref target="RFC8693"/>, rich authorization requests <xref target="RFC9396"/>, DPoP <xref target="RFC9449"/></td><td>Authorize and bind callers and requests; they do not reconstruct the operation at the sink or consume authority per effect.</td></tr>
<tr><td>RATS <xref target="RFC9334"/>, EAT <xref target="RFC9711"/>, TPM, HSM, Secure Enclave, Android Keystore, Arm CCA, Intel TDX, AMD SEV-SNP, secure and measured boot</td><td>Attest environments and hold keys; attestation shows the environment, not each effect.</td></tr>
<tr><td>Android, Apple, Windows and Linux security models</td><td>Isolate and authorize processes; no cross-platform effect semantics.</td></tr>
<tr><td>MCP, A2A, agent SDKs, tool gateways, service meshes, policy enforcement points</td><td>Orchestrate and gate calls; coverage and operation semantics vary.</td></tr>
<tr><td>Fencing tokens, leases, generation counters, idempotency keys, journals, outbox, consensus, sagas</td><td>Known distributed-systems techniques reused here; none is claimed as new in isolation.</td></tr>
</tbody>
</table>
<t>The examiner discussion (objective technical problem, motivation to combine, reasonable expectation of success, functional interdependence, hindsight) is in <xref target="app-f"/> and is non-normative.</t>
</section>
<section anchor="iana">
<name>IANA Considerations</name>
<t>This document has no IANA actions.</t>
</section>
<section anchor="ipr">
<name>Intellectual Property Rights</name>
<t>Human inventor: Sangam Kumar Das. Current patent source: Indian provisional application identified by the applicant as 202631127058, reported filed 25 October 2026, technical title as stated in the note at the start. The source is a 3,577-page PDF with 40 distinct numbered embodiments and 60 formal claims. These details are applicant-identified and must be checked against the official acknowledgment.</t>
<t>Foundational application: PCT/IB2026/055615, filed 4 June 2026, published 16 July 2026 as WO 2026/150382 A2 (<xref target="PCT-055615"/>; WIPO PATENTSCOPE https://patentscope.wipo.int/). The 25 October filing is part of the earlier human-invented DAS Protocols research programme; its cross-platform mechanisms are to be differentiated from the foundational architecture and assessed on their specific features. Related public Internet-Drafts are <xref target="REL-FOUND"/>, <xref target="REL-IEV"/> and <xref target="REL-CBEA"/>; newer revisions were not checked. Publication on the Datatracker is not adoption, patent examination or validation.</t>
<t>The IPR disclosure for this document will be made at the appropriate time in accordance with IETF guidelines and BCP 78 and BCP 79 (<xref target="RFC8179"/>). A patent statement inside a draft does not replace an official IETF IPR disclosure. No royalty-free, FRAND or other licensing commitment is made here.</t>
</section>
</middle>
<back>
<references><name>Normative References</name>
<reference anchor="RFC2119" target="https://www.rfc-editor.org/info/rfc2119"><front><title>Key words for use in RFCs to Indicate Requirement Levels</title><author initials="S" surname="Bradner"/><date year="1997" month="March"/></front><seriesInfo name="RFC" value="2119"/><seriesInfo name="BCP" value="14"/></reference>
<reference anchor="RFC8174" target="https://www.rfc-editor.org/info/rfc8174"><front><title>Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words</title><author initials="B" surname="Leiba"/><date year="2017" month="May"/></front><seriesInfo name="RFC" value="8174"/><seriesInfo name="BCP" value="14"/></reference>
</references>
<references><name>Informative References</name>
<reference anchor="RFC3552" target="https://www.rfc-editor.org/info/rfc3552"><front><title>Guidelines for Writing RFC Text on Security Considerations</title><author initials="E" surname="Rescorla"/><author initials="B" surname="Korver"/><date year="2003" month="July"/></front><seriesInfo name="RFC" value="3552"/></reference>
<reference anchor="RFC6749" target="https://www.rfc-editor.org/info/rfc6749"><front><title>The OAuth 2.0 Authorization Framework</title><author initials="D" surname="Hardt"/><date year="2012" month="October"/></front><seriesInfo name="RFC" value="6749"/></reference>
<reference anchor="RFC6973" target="https://www.rfc-editor.org/info/rfc6973"><front><title>Privacy Considerations for Internet Protocols</title><author initials="A" surname="Cooper"/><author initials="H" surname="Tschofenig"/><author initials="B" surname="Aboba"/><author initials="J" surname="Peterson"/><author initials="J" surname="Morris"/><author initials="M" surname="Hansen"/><author initials="R" surname="Smith"/><date year="2013" month="July"/></front><seriesInfo name="RFC" value="6973"/></reference>
<reference anchor="RFC8179" target="https://www.rfc-editor.org/info/rfc8179"><front><title>Intellectual Property Rights in IETF Technology</title><author initials="S" surname="Bradner"/><author initials="J" surname="Contreras"/><date year="2017" month="May"/></front><seriesInfo name="RFC" value="8179"/><seriesInfo name="BCP" value="79"/></reference>
<reference anchor="RFC8693" target="https://www.rfc-editor.org/info/rfc8693"><front><title>OAuth 2.0 Token Exchange</title><author initials="M" surname="Jones"/><author initials="A" surname="Nadalin"/><author initials="B" surname="Campbell"/><author initials="J" surname="Bradley"/><author initials="C" surname="Mortimore"/><date year="2020" month="January"/></front><seriesInfo name="RFC" value="8693"/></reference>
<reference anchor="RFC8949" target="https://www.rfc-editor.org/info/rfc8949"><front><title>Concise Binary Object Representation (CBOR)</title><author initials="C" surname="Bormann"/><author initials="P" surname="Hoffman"/><date year="2020" month="December"/></front><seriesInfo name="RFC" value="8949"/></reference>
<reference anchor="RFC9334" target="https://www.rfc-editor.org/info/rfc9334"><front><title>Remote ATtestation procedureS (RATS) Architecture</title><author initials="H" surname="Birkholz"/><author initials="D" surname="Thaler"/><author initials="M" surname="Richardson"/><author initials="N" surname="Smith"/><author initials="W" surname="Pan"/><date year="2023" month="January"/></front><seriesInfo name="RFC" value="9334"/></reference>
<reference anchor="RFC9396" target="https://www.rfc-editor.org/info/rfc9396"><front><title>OAuth 2.0 Rich Authorization Requests</title><author initials="T" surname="Lodderstedt"/><author initials="J" surname="Richer"/><author initials="B" surname="Campbell"/><date year="2023" month="May"/></front><seriesInfo name="RFC" value="9396"/></reference>
<reference anchor="RFC9449" target="https://www.rfc-editor.org/info/rfc9449"><front><title>OAuth 2.0 Demonstrating Proof of Possession (DPoP)</title><author initials="D" surname="Fett"/><author initials="B" surname="Campbell"/><author initials="J" surname="Bradley"/><author initials="T" surname="Lodderstedt"/><author initials="M" surname="Jones"/><author initials="D" surname="Waite"/><date year="2023" month="September"/></front><seriesInfo name="RFC" value="9449"/></reference>
<reference anchor="RFC9711" target="https://www.rfc-editor.org/info/rfc9711"><front><title>The Entity Attestation Token (EAT)</title><author initials="L" surname="Lundblad"/><author initials="G" surname="Mandyam"/><author initials="J" surname="O'Donoghue"/><author initials="C" surname="Wallace"/><date year="2025" month="April"/></front><seriesInfo name="RFC" value="9711"/></reference>
<reference anchor="REL-FOUND" target="https://datatracker.ietf.org/"><front><title>Execution Finality Foundations</title><author fullname="Sangam Kumar Das" initials="S." surname="Das"/><date year="2026"/></front><seriesInfo name="Internet-Draft" value="draft-das-execution-finality-foundations"/><refcontent>Identifier as supplied; verify on the Datatracker. Not fetched.</refcontent></reference>
<reference anchor="REL-AGENT" target="https://datatracker.ietf.org/"><front><title>Agent Execution Finality (title as supplied; verify)</title><author fullname="Sangam Kumar Das" initials="S." surname="Das"/><date year="2026"/></front><seriesInfo name="Internet-Draft" value="draft-das-agent-execution-finality"/><refcontent>Identifier as supplied; verify on the Datatracker. Not fetched.</refcontent></reference>
<reference anchor="REL-IEV" target="https://datatracker.ietf.org/"><front><title>Receipt-Gated Independent Effect Verification (title as supplied; verify)</title><author fullname="Sangam Kumar Das" initials="S." surname="Das"/><date year="2026"/></front><seriesInfo name="Internet-Draft" value="draft-das-receipt-gated-iev"/><refcontent>Identifier as supplied; verify on the Datatracker. Not fetched.</refcontent></reference>
<reference anchor="REL-CBEA" target="https://datatracker.ietf.org/"><front><title>Consequence-Bound Execution Authority</title><author fullname="Sangam Kumar Das" initials="S." surname="Das"/><date year="2026"/></front><seriesInfo name="Internet-Draft" value="draft-das-consequence-bound-execution-authority"/><refcontent>Identifier as supplied; verify on the Datatracker. Not fetched.</refcontent></reference>
<reference anchor="PCT-055615" target="https://patentscope.wipo.int/"><front><title>PCT/IB2026/055615 (WO 2026/150382 A2)</title><author fullname="Sangam Kumar Das" initials="S." surname="Das"/><date year="2026" month="July" day="16"/></front><refcontent>Filed 4 June 2026; published 16 July 2026. Verify on PATENTSCOPE.</refcontent></reference>
<reference anchor="TDC-12542" target="https://www.tdcommons.org/dpubs_series/12542/"><front><title>Technical Disclosure Commons record 12542 (PCT file; title not verified)</title><author fullname="Sangam Das" initials="S." surname="Das"/><date year="2026"/></front><refcontent>Not fetched; verify.</refcontent></reference>
<reference anchor="IN-127058" target="https://ipindia.gov.in/"><front><title>Indian provisional application 202631127058 (applicant-identified)</title><author fullname="Sangam Kumar Das" initials="S." surname="Das"/><date year="2026" month="October" day="25"/></front><refcontent>Applicant-identified; not a certified registry record.</refcontent></reference>
</references>
<section anchor="app-a">
<name>Appendix A: Embodiments E1-E40</name>
<t>The 40 distinct embodiments follow the source's numbering. Start pages are pages of the supplied 3,577-page PDF, approximate where the source repeats headings (E1 and E2 headings recur near pages 364 and 378). The second full section labelled Embodiment 31 (about pages 2650-2734) is a repeated version and is not counted as an embodiment. Clusters A-F are those of <xref target="clusters"/>; an embodiment may belong to more than one cluster.</t>
<table>
<thead><tr><th>No.</th><th>Technical subject</th><th>Cluster</th><th>Start page</th><th>Workflow</th></tr></thead>
<tbody>
<tr><td>E1</td><td>Protected cross-agent effect-conflict arbitration with atomic reservations and sink-bound fencing</td><td>A</td><td>248</td><td>CW-1</td></tr>
<tr><td>E2</td><td>Identity-bound Interim Effectuation Validation and receipt-constituted causal continuation</td><td>B</td><td>295</td><td>CW-1, CW-2</td></tr>
<tr><td>E3</td><td>User-referenced semantic object correspondence and authenticated presentation-state binding</td><td>C</td><td>363</td><td>CW-1</td></tr>
<tr><td>E4</td><td>Cross-domain actual-consequence witnessing and qualified remote-effect evidence</td><td>B</td><td>492</td><td>CW-1</td></tr>
<tr><td>E5</td><td>Cryptographically verifiable, assistant-neutral resource allocation and protected eligible-demand completeness</td><td>A</td><td>555</td><td>CW-1</td></tr>
<tr><td>E6</td><td>Semantic action-contract evolution and effect-compatibility validation</td><td>C</td><td>666</td><td>CW-1</td></tr>
<tr><td>E7</td><td>Cross-provider delegated agent authority continuity and cryptographic delegation lineage</td><td>D</td><td>723</td><td>CW-1</td></tr>
<tr><td>E8</td><td>Correlation-aware cumulative privacy exposure and compositional disclosure constraints</td><td>C</td><td>787</td><td>CW-1</td></tr>
<tr><td>E9</td><td>Cross-device execution-authority ownership, hardware epochs and source fencing</td><td>D</td><td>859</td><td>CW-1</td></tr>
<tr><td>E10</td><td>Model, algorithm and runtime-generation drift with effectuation-time revalidation</td><td>E</td><td>926</td><td>CW-1</td></tr>
<tr><td>E11</td><td>Concurrent-effect reconciliation, version-bound reservations and sink-local commit validation</td><td>A</td><td>994</td><td>CW-1</td></tr>
<tr><td>E12</td><td>Autonomous recovery, compensating effects and non-replayable recovery authority</td><td>F</td><td>1063</td><td>CW-1</td></tr>
<tr><td>E13</td><td>Multi-agent consensus, quorum integrity and non-delegable finality authorization</td><td>A</td><td>1140</td><td>CW-1</td></tr>
<tr><td>E14</td><td>Effectuation-path discovery, bypass detection and enforcement-coverage revalidation</td><td>E</td><td>1221</td><td>CW-1</td></tr>
<tr><td>E15</td><td>Cross-domain evidence commitments, non-bearer authority and sink-local monotonic consumption</td><td>B</td><td>1309</td><td>CW-1</td></tr>
<tr><td>E16</td><td>Crash-consistent finality journaling, one-time authorization and indeterminate outcome handling</td><td>B,F</td><td>1384</td><td>CW-1</td></tr>
<tr><td>E17</td><td>Temporal authority, lease expiration, clock uncertainty and delayed-effect control</td><td>D</td><td>1465</td><td>CW-1</td></tr>
<tr><td>E18</td><td>Cross-domain revocation propagation and dependent authority invalidation</td><td>D,F</td><td>1541</td><td>CW-1</td></tr>
<tr><td>E19</td><td>Computer-use and transaction-bound human-approval finality</td><td>C</td><td>1625</td><td>CW-1</td></tr>
<tr><td>E20</td><td>Multi-step tool-chain and workflow dependency-bound execution authority</td><td>E</td><td>1683</td><td>CW-1</td></tr>
<tr><td>E21</td><td>Actual resource identity, aliases, redirects and dynamic namespace rebinding</td><td>C</td><td>1758</td><td>CW-1</td></tr>
<tr><td>E22</td><td>Multi-resource budgets, delegated quotas and cumulative consequential limits</td><td>A</td><td>1830</td><td>CW-1</td></tr>
<tr><td>E23</td><td>Credential and secret-use finality with non-exportable broker-controlled authority</td><td>E</td><td>1897</td><td>CW-1</td></tr>
<tr><td>E24</td><td>Tool-output and instruction provenance through model-driven execution</td><td>C</td><td>2019</td><td>CW-1</td></tr>
<tr><td>E25</td><td>Agent sandbox privilege transitions and consequential operations escaping sandbox boundaries</td><td>E</td><td>2092</td><td>CW-1</td></tr>
<tr><td>E26</td><td>Tool installation, updates and runtime lifecycle authority continuity</td><td>E</td><td>2169</td><td>CW-1</td></tr>
<tr><td>E27</td><td>Distributed agent handoff, workload migration and non-resurrectable in-flight effects</td><td>D,F</td><td>2246</td><td>CW-1</td></tr>
<tr><td>E28</td><td>Nested multi-agent tool invocation chains and transitive authority attenuation</td><td>E</td><td>2318</td><td>CW-1</td></tr>
<tr><td>E29</td><td>Independent approval quorums and protected consensus-to-effect separation</td><td>A</td><td>2398</td><td>CW-1</td></tr>
<tr><td>E30</td><td>Resource-owner changes, competing administrative authority and revocation races</td><td>D</td><td>2480</td><td>CW-1</td></tr>
<tr><td>E31</td><td>Credential and administrative authority rotation, protected key custody and generation fencing</td><td>D</td><td>2567</td><td>CW-1</td></tr>
<tr><td>E32</td><td>Data, memory and persistent-knowledge provenance across agent sessions</td><td>C</td><td>2734</td><td>CW-1</td></tr>
<tr><td>E33</td><td>Policy and configuration evolution, dependency impact and non-self-authorizing mutations</td><td>E</td><td>2825</td><td>CW-1</td></tr>
<tr><td>E34</td><td>Tool and resource identity continuity across dynamic service discovery</td><td>C</td><td>2914</td><td>CW-1</td></tr>
<tr><td>E35</td><td>Degraded connectivity, offline effect authority and delayed revocation</td><td>D,F</td><td>3005</td><td>CW-1</td></tr>
<tr><td>E36</td><td>Multi-resource transactions, partial commits and independently authorized compensation</td><td>F</td><td>3102</td><td>CW-1</td></tr>
<tr><td>E37</td><td>Irreversible effects, consequence-bound hazard checks and non-compensable operation containment</td><td>F</td><td>3192</td><td>CW-1</td></tr>
<tr><td>E38</td><td>Evidence integrity, compromised observers and independent attestation reconstruction</td><td>B</td><td>3282</td><td>CW-1</td></tr>
<tr><td>E39</td><td>Conflicting policies and authority precedence among multiple principals and resource owners</td><td>D</td><td>3378</td><td>CW-1</td></tr>
<tr><td>E40</td><td>End-to-end effect provenance, protected authority consumption, cross-embodiment closure and formal verification</td><td>B</td><td>3474</td><td>CW-1</td></tr>
</tbody>
</table>
<t>Source statements of the principal candidate inventive concept (excerpts, lightly normalised; extracted automatically, so check against the PDF):</t>
<ul>
<li>E5: the non-bypassable transformation of a cryptographically verifiable, policy-neutral resource-allocation decision— computed from authenticated eligible demand—into scoped resource-grant authority independently verified at the actual resource Finality Sink, together with protected evidence linking the authorized grant to actual ...</li>
<li>E8: the non-bypassable enforcement of cumulative, recipient-scoped, source-dependent privacy constraints at an actual information-release Finality Sink, wherein the protected system evaluates a new disclosure against relevant previous releases across applications or autonomous agents, atomically reserves the permissible exposure ...</li>
<li>E9: a protected cross-device authority-transfer architecture in which an unfinished consequential operation remains subject to authenticated exclusive ownership, qualified source- effectuation fencing, authoritative generation control, and destination-specific non- bearer authority, with the actual destination Finality Sink ...</li>
<li>E11: a protected concurrent-effect finality architecture in which authenticated resource-version and semantic conflict dependencies are bound to operation-specific protected reservations, and a non-bypassable Finality Sink independently reconstructs the actual pending consequential operation and refuses effectuation unless its ...</li>
<li>E13: a protected multi-agent collective authorization architecture in which an originating Candidate Act is bound to a canonical consequential operation, authenticated participants satisfy a current protected quorum and role policy, qualified approval evidence supports scoped non-bearer finality authority, and an independently ...</li>
<li>E14: a protected effectuation-path continuity architecture in which authenticated effect-capable topology and blocking-enforcement evidence support a coverage-bound finality authority, material runtime or hardware path changes invalidate affected protected authority, and an independently enforcing Finality Sink refuses the actual ...</li>
<li>E15: a protected cross-domain evidence-to-effectuation architecture in which collection-time evidence from a first protected instance is independently validated by a second protected instance, the respective evidence states are cross-committed to one specified Candidate Act, scoped non-bearer finality authority is established ...</li>
<li>E16: a crash- consistent protected execution-finality system in which an independently enforcing Finality Sink reconstructs an actual pending consequential operation, verifies current operation-specific non-bearer authority, durably binds a monotonic authority- consumption transition to the corresponding actual-effect ...</li>
<li>E17: bound to an authenticated operation-specific temporal authority envelope, a protected verifier establishes qualified current time evidence under applicable uncertainty and lease-generation conditions, a scoped non-bearer finality enablement remains dependent upon those current conditions, and an independent Finality Sink ...</li>
<li>E18: a protected cross- domain revocation-finality system in which an authenticated source-authority revocation is committed to monotonic protected state, an authority-dependency controller identifies and invalidates affected delegated and operation-specific non-bearer finality enablements, participating resource domains establish ...</li>
<li>E19: a protected computer-use execution-finality system in which qualified interface-state evidence and an operation-specific authorized-effect descriptor are bound to a Candidate Act, optional authenticated human confirmation is committed to the same material effect parameters, and an independent Finality Sink reconstructs the ...</li>
<li>E20: a protected autonomous workflow system in which a machine-verifiable dependency mechanism distinguishes computational progress from qualified actual-effect evidence, a protected continuation controller independently establishes current authority and required predecessor predicates for each consequential successor Candidate Act, ...</li>
<li>E21: a protected autonomous resource-identity execution-finality system in which the original source- authorized resource referent and a qualified current resolution relationship are bound to operation-specific non-bearer finality authority, and an independently enforcing Finality Sink reconstructs the actual pending resource target ...</li>
<li>E22: a protected multidimensional autonomous resource-finality architecture in which source-authorized effect budgets are delegated through non-overlapping protected allocations, consequential Candidate Acts receive operation-specific reserved authority and separate non-bearer finality enablement, an independently enforcing Finality ...</li>
<li>E23: a protected autonomous credential-use system in which reusable credential material remains under qualified protected custody, a consequential Candidate Act is independently validated against current originating authority, a separate scoped non-bearer finality enablement is constituted, and a credential-use Finality Sink ...</li>
<li>E24: a protected autonomous execution-finality system in which current originating instruction authority is maintained independently from lower-trust informational provenance and computational derivations, a proposed consequential operation is represented as a non- effective Candidate Act, protected semantic verification establishes ...</li>
<li>E25: a protected autonomous sandbox-to-effectuation architecture in which a lower-trust computational workload remains confined within a qualified execution domain, operations crossing a consequential boundary are separately represented and withheld as Candidate Acts, required privilege transitions are constituted as bounded ...</li>
<li>E26: a protected autonomous tool-lifecycle execution-finality system in which pending consequential Candidate Acts are bound to qualified effect-relevant tool and runtime state, material lifecycle changes cause protected selective fencing and current revalidation, operation- specific non-bearer finality is reconstituted where ...</li>
<li>E27: a protected distributed autonomous execution-finality system in which operation-specific source authority is preserved across workload migration through source-domain fencing, qualified destination handoff, monotonic effectuation-ownership generations, independent reconciliation of in-flight real effects, destination-specific ...</li>
<li>E28: a protected autonomous multi-agent execution-finality system in which transitive source-authority restrictions remain independently verifiable through nested and branching delegation relationships, shared consequential allowances are conserved across concurrent descendants, intermediate resource effects are held as ...</li>
<li>E29: a protected autonomous consensus-to-execution-finality architecture in which qualified multi-agent approval is cryptographically and semantically bound to one Candidate Act, conflict and veto predicates are resolved under current protected quorum policy, source and resource authority remain independently mandatory, and an ...</li>
<li>E30: a protected resource-authority-continuity system in which a consequential Candidate Act is associated with authenticated operation-specific source and resource authority dependencies, material ownership and revocation transitions selectively fence pending finality, competing administrative authority claims are resolved ...</li>
<li>E31: a protected autonomous credential-dependent execution-finality system in which reusable credential custody is separated from ordinary agent computation, the actual Candidate Act is bound to credential generation and source-authority dependencies, material credential transitions trigger protected finality fencing and current ...</li>
<li>E32: a protected autonomous memory-to-effect-finality system in which persistent information is governed by independent source-provenance and storage-finality controls, cross-session retrieval is conditioned upon current authority without promoting retrieved data into independent instruction authority, and consequential operations ...</li>
<li>E33: a protected autonomous policy-to-effect-finality system in which policy changes are separately non- effective Candidate Acts governed by competent meta-authority, qualified semantic impact and dependency analysis selectively fence affected pending operations, actual policy activation is independently enforced, and subsequent ...</li>
<li>E34: a protected autonomous resource-identity-to-execution-finality system in which a held Candidate Act is associated with authorized actual resource and tool-effect conditions, dynamic discovery and tool substitution are qualified under protected identity-continuity rules, material changes fence pending finality, and an ...</li>
<li>E35: a protected autonomous partition-aware execution-finality system in which source-authorized consequential operations are classified by current connectivity requirements, selected prepartition authority is attenuated into bounded sink-specific offline allocations, actual local effects remain non-effective until independently ...</li>
<li>E36: a protected autonomous multi-resource transaction-finality system in which source-authorized transactions are decomposed into independently non-effective constituent effects, protected actual-effect receipts control causal successor release without expanding source authority, partial-commit containment preserves completed and ...</li>
<li>E37: a protected autonomous consequence-to-effectuation-finality system in which a source-authorized irreversible operation remains non-effective until the actual effectuation frontier is identified, current hazard and resource predicates are independently qualified, material changes fence pending authority, and a scoped non-bearer ...</li>
<li>E38: a protected autonomous evidence-integrity execution-finality system in which a source-authorized Candidate Act remains non-effective while collection-time and independently acquired execution-time observations are qualified for observer identity, competence, continuity, independence, and actual effect correspondence; ...</li>
<li>E39: an autonomous multi-authority execution-finality system in which source-authorized consequential Candidate Acts remain non-effective while competent operation-specific authority provenance, delegation, restrictions, policy precedence, and conflict conditions are qualified; unresolved mandatory authority fences effectuation; ...</li>
</ul>
<t>Embodiments without an extracted statement: E1, E2, E3, E4, E6, E7, E10, E12, E40. The requested 20-point analysis (objective, components, authority, candidate operation, effect-critical data, protected state, enforcement interface, predicate, inputs and outputs, evidence, attack scenarios, negative outcomes, recovery, test method, performance, existing foundations, narrower features, industry integration, workflow relationship, mothership relationship) was done at cluster level in <xref target="clusters"/> and in the threat, example and industry appendices, not individually for each of the 40 embodiments; per-embodiment pseudocode and data structures from the source were not extracted.</t>
</section>
<section anchor="app-b">
<name>Appendix B: Claims 1-60</name>
<t>Non-normative. The claims are not requirements of this document. The consolidated claims contain four independent claims: 1 (system, dependents 2-30), 31 (method, 32-42), 43 (apparatus, 44-52) and 53 (migration, 54-60). Forty embodiments do not correspond to forty independent claims, and no missing dependent claims were invented. Independent versus dependent was determined from the claim text, and the dependency shown is the first claim number referenced. Excerpts are from the extracted text (ligatures repaired automatically); check them against the PDF.</t>
<table>
<thead><tr><th>No.</th><th>Type</th><th>Dep.</th><th>Group</th><th>Source limitation (excerpt)</th><th>Realization wording (whole claim)</th></tr></thead>
<tbody>
<tr><td>1</td><td>Independent</td><td>-</td><td>System</td><td>A computer-implemented system for interoperable protected execution-finality control across heterogeneous computing environments, the system comprising: (a) a first ...</td><td>none named in the claim text</td></tr>
<tr><td>2</td><td>Dependent</td><td>1</td><td>System</td><td>the first and second computing environments comprise different operating systems, different administrative domains, different device-security architectures, or different ...</td><td>none named in the claim text</td></tr>
<tr><td>3</td><td>Dependent</td><td>2</td><td>System</td><td>the first computing environment comprises an Android-based device and the second computing environment comprises an iOS-based device, and wherein a consequential ...</td><td>none named in the claim text</td></tr>
<tr><td>4</td><td>Dependent</td><td>2</td><td>System</td><td>the first computing environment comprises an iOS- based device and the second computing environment comprises an Android-based device, and wherein the second computing ...</td><td>none named in the claim text</td></tr>
<tr><td>5</td><td>Dependent</td><td>2</td><td>System</td><td>the heterogeneous computing environments comprise any compatible combination of mobile operating systems, desktop operating systems, virtualized execution environments, ...</td><td>none named in the claim text</td></tr>
<tr><td>6</td><td>Dependent</td><td>1</td><td>System</td><td>the heterogeneous authorization-source adapter comprises a typed interface identifying a technical security property established by a corresponding underlying mechanism, ...</td><td>none named in the claim text</td></tr>
<tr><td>7</td><td>Dependent</td><td>1</td><td>System</td><td>the qualified evidence comprises one or more of an operating-system permission, application entitlement, source authorization grant, structured authorization detail, ...</td><td>cryptographic terms</td></tr>
<tr><td>8</td><td>Dependent</td><td>1</td><td>System</td><td>the protected canonical effect descriptor comprises a selected combination of a Candidate Act identifier, originating authority reference, actual operation class, actual ...</td><td>none named in the claim text</td></tr>
<tr><td>9</td><td>Dependent</td><td>1</td><td>System</td><td>the protected authorization-to-effect translation controller resolves the operation semantics according to an authenticated action contract or resource-interface ...</td><td>none named in the claim text</td></tr>
<tr><td>10</td><td>Dependent</td><td>1</td><td>System</td><td>the Protected Enforcement Domain maintains a protected relationship between an originating source grant and one or more dependent delegated authorizations, Candidate ...</td><td>none named in the claim text</td></tr>
<tr><td>11</td><td>Dependent</td><td>1</td><td>System</td><td>the protected validation evidence comprises collection-time evidence established by a first protected validation instance and execution-time evidence established by a ...</td><td>none named in the claim text</td></tr>
<tr><td>12</td><td>Dependent</td><td>1</td><td>System</td><td>a structured application action, application-intent object, agent tool invocation, interprocess service request, or remote application-function call is represented as ...</td><td>none named in the claim text</td></tr>
<tr><td>13</td><td>Dependent</td><td>1</td><td>System</td><td>the Non-Effective Holding Mechanism comprises one or more of a blocked application-function dispatch, withheld network transmission, uncommitted storage operation, ...</td><td>none named in the claim text</td></tr>
<tr><td>14</td><td>Dependent</td><td>1</td><td>System</td><td>a proposed application action or autonomous workflow comprises multiple distinguishable consequential effects, and wherein at least two of the consequential effects are ...</td><td>none named in the claim text</td></tr>
<tr><td>15</td><td>Dependent</td><td>1</td><td>System</td><td>The system as claimed in claim 1, further comprising an Enforcement-Coverage Qualification Controller configured to maintain a protected enforcement-qualification record ...</td><td>none named in the claim text</td></tr>
<tr><td>16</td><td>Dependent</td><td>15</td><td>System</td><td>an observation-only security monitor, notification interface, advisory security policy, or post-effect audit mechanism is prevented from satisfying a fl mandatory ...</td><td>none named in the claim text</td></tr>
<tr><td>17</td><td>Dependent</td><td>15</td><td>System</td><td>the Enforcement-Coverage Qualification Controller identifies a plurality of interfaces capable of producing a governed effect and requires each interface within an ...</td><td>none named in the claim text</td></tr>
<tr><td>18</td><td>Dependent</td><td>15</td><td>System</td><td>a material change in an operating-system security interface, sandbox entitlement, application-dispatch configuration, privileged intermediary, network route, ...</td><td>hardware terms</td></tr>
<tr><td>19</td><td>Dependent</td><td>1</td><td>System</td><td>the Protected Enforcement Domain maintains a current revocation generation associated with an originating authorization grant and identifies scoped finality enablements ...</td><td>none named in the claim text</td></tr>
<tr><td>20</td><td>Dependent</td><td>19</td><td>System</td><td>the designated Finality Sink verifies an applicable current revocation generation or protected authorization-status condition, and wherein an earlier active-token ...</td><td>none named in the claim text</td></tr>
<tr><td>21</td><td>Dependent</td><td>1</td><td>System</td><td>the scoped non-bearer finality enablement is maintained within a protected controller or sink-local protected state and is represented to the requesting computational ...</td><td>none named in the claim text</td></tr>
<tr><td>22</td><td>Dependent</td><td>21</td><td>System</td><td>the scoped non-bearer finality enablement comprises a permitted-use count or unique consumption reference, and wherein the designated Finality Sink enforces a protected ...</td><td>none named in the claim text</td></tr>
<tr><td>23</td><td>Dependent</td><td>21</td><td>System</td><td>the scoped non-bearer finality enablement is bound to one or more of the designated Finality Sink, actual resource, permitted effect, originating authority, validity ...</td><td>none named in the claim text</td></tr>
<tr><td>24</td><td>Dependent</td><td>1</td><td>System</td><td>the designated Finality Sink determines at least one security-relevant descriptor element from the actual operation presented at the controlled effectuation interface ...</td><td>none named in the claim text</td></tr>
<tr><td>25</td><td>Dependent</td><td>24</td><td>System</td><td>the designated Finality Sink verifies one or more of an actual destination, resource identity, recipient, outgoing payload commitment, application- function parameter, ...</td><td>none named in the claim text</td></tr>
<tr><td>26</td><td>Dependent</td><td>1</td><td>System</td><td>the designated Finality Sink comprises a transactional resource controller configured to atomically commit, within an applicable transaction boundary, a protected ...</td><td>none named in the claim text</td></tr>
<tr><td>27</td><td>Dependent</td><td>1</td><td>System</td><td>The system as claimed in claim 1, further comprising a protected credential broker configured to retain a reusable resource credential outside the requesting ...</td><td>none named in the claim text</td></tr>
<tr><td>28</td><td>Dependent</td><td>1</td><td>System</td><td>a governed consequential operation comprises transmission of information through a network interface, and wherein protected authorization evidence binds the permitted ...</td><td>none named in the claim text</td></tr>
<tr><td>29</td><td>Dependent</td><td>1</td><td>System</td><td>the first computing environment, the Protected Enforcement Domain, and the designated Finality Sink reside in at least two independently administered domains, and ...</td><td>none named in the claim text</td></tr>
<tr><td>30</td><td>Dependent</td><td>1</td><td>System</td><td>a computing environment operates with intermittent connectivity and the Protected Enforcement Domain establishes a bounded of ine authority comprising one or more ...</td><td>none named in the claim text</td></tr>
<tr><td>31</td><td>Independent</td><td>-</td><td>Method</td><td>A computer-implemented method for enforcing protected execution-finality control across heterogeneous computing environments, the method comprising: (a) receiving, from ...</td><td>none named in the claim text</td></tr>
<tr><td>32</td><td>Dependent</td><td>31</td><td>Method</td><td>obtaining qualified evidence and independently validating the proposed consequential operation comprise collecting authenticated evidence at a first protected ...</td><td>none named in the claim text</td></tr>
<tr><td>33</td><td>Dependent</td><td>31</td><td>Method</td><td>producing the protected canonical effect descriptor comprises converting heterogeneous native-platform permission or authorization information into a typed effect ...</td><td>none named in the claim text</td></tr>
<tr><td>34</td><td>Dependent</td><td>31</td><td>Method</td><td>The method as claimed in claim 31, further comprising detecting a material change in resource identity, resource generation, effective recipient, operation parameters, ...</td><td>none named in the claim text</td></tr>
<tr><td>35</td><td>Dependent</td><td>31</td><td>Method</td><td>The method as claimed in claim 31, further comprising obtaining an authenticated revocation event or current revoked-state indication, identifying pending finality ...</td><td>none named in the claim text</td></tr>
<tr><td>36</td><td>Dependent</td><td>31</td><td>Method</td><td>the designated Finality Sink comprises a qualified protected application dispatcher, privileged interprocess broker, network gateway, storage controller, remote service ...</td><td>hardware terms</td></tr>
<tr><td>37</td><td>Dependent</td><td>31</td><td>Method</td><td>successful user approval, application authentication, cryptographic signature verification, credential introspection, or operating-system permission validation is ...</td><td>cryptographic terms</td></tr>
<tr><td>38</td><td>Dependent</td><td>31</td><td>Method</td><td>original source authority, delegated authority, protected evidence, or finality-enablement state is maintained in a protected storage or controller domain not modifiable ...</td><td>none named in the claim text</td></tr>
<tr><td>39</td><td>Dependent</td><td>31</td><td>Method</td><td>The method as claimed in claim 31, further comprising revalidating the actual pending operation at the designated Finality Sink using current protected resource or ...</td><td>none named in the claim text</td></tr>
<tr><td>40</td><td>Dependent</td><td>31</td><td>Method</td><td>an unsuccessful or unavailable mandatory validator, invalid finality enablement, unresolved operation semantics, stale authority generation, or unqualified effectuation ...</td><td>none named in the claim text</td></tr>
<tr><td>41</td><td>Dependent</td><td>31</td><td>Method</td><td>absence of an acknowledgment or an uncertain remote result does not automatically restore consumed authority or authorize an unrestricted retry of the operation.</td><td>none named in the claim text</td></tr>
<tr><td>42</td><td>Dependent</td><td>41</td><td>Method</td><td>a successor consequential operation depends upon an earlier real effect and is withheld until an authenticated receipt concerning the earlier effect has been accepted by ...</td><td>none named in the claim text</td></tr>
<tr><td>43</td><td>Independent</td><td>-</td><td>Apparatus</td><td>A protected computing apparatus for enforcing interoperable execution-finality authorization at a consequential resource boundary, the apparatus comprising: (a) at least ...</td><td>none named in the claim text</td></tr>
<tr><td>44</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the apparatus comprises a mobile computing device, desktop computer, server, protected application-execution component, embedded controller, secure processor, cloud ...</td><td>none named in the claim text</td></tr>
<tr><td>45</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the apparatus is associated with an iOS-based or iPadOS-based computing environment and the resource interface comprises an authorized application function, protected ...</td><td>none named in the claim text</td></tr>
<tr><td>46</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the apparatus is associated with an Android- based computing environment and the resource interface comprises a protected application dispatcher, privileged service, ...</td><td>hardware terms</td></tr>
<tr><td>47</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the resource interface comprises an authenticated interprocess communication broker separately controlled from the requesting computational entity, and wherein the ...</td><td>none named in the claim text</td></tr>
<tr><td>48</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the apparatus implements destination-side protected enforcement for a remote application or service, thereby permitting interoperability with a requesting operating ...</td><td>none named in the claim text</td></tr>
<tr><td>49</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the effectuation controller comprises a hardware- supported command gate, protected storage controller, secure processor, device command controller, or hardware-backed ...</td><td>hardware terms</td></tr>
<tr><td>50</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the interoperable authority-verification component maintains a protected enforcement-qualification record and rejects a claimed mandatory enforcement property when an ...</td><td>none named in the claim text</td></tr>
<tr><td>51</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the canonical effect descriptor is represented using a typed structured encoding or deterministic binary representation de ning canonicalization rules for ...</td><td>none named in the claim text</td></tr>
<tr><td>52</td><td>Dependent</td><td>43</td><td>Apparatus</td><td>the interoperable authority-verification component verifies authenticated principal attribution, originating task authority, and permitted effect scope independently of ...</td><td>none named in the claim text</td></tr>
<tr><td>53</td><td>Independent</td><td>-</td><td>Migration</td><td>A computer-implemented system for protected migration of consequential execution authority between heterogeneous computing environments, the system comprising: (a) a ...</td><td>none named in the claim text</td></tr>
<tr><td>54</td><td>Dependent</td><td>53</td><td>Migration</td><td>the protected authority-ownership controller maintains a monotonic ownership epoch, protected generation counter, authenticated fencing reference, or authoritative ...</td><td>none named in the claim text</td></tr>
<tr><td>55</td><td>Dependent</td><td>53</td><td>Migration</td><td>verification of the source-fencing condition comprises verifying restriction, invalidation, or isolation of at least one effect-capable source-device interface, pending ...</td><td>hardware terms</td></tr>
<tr><td>56</td><td>Dependent</td><td>53</td><td>Migration</td><td>the source and destination execution environments comprise respectively different mobile or desktop operating systems, including an Android-based environment and an ...</td><td>none named in the claim text</td></tr>
<tr><td>57</td><td>Dependent</td><td>53</td><td>Migration</td><td>loss of communication with the source execution environment does not automatically establish successful source fencing, and wherein activation of exclusive destination ...</td><td>none named in the claim text</td></tr>
<tr><td>58</td><td>Dependent</td><td>53</td><td>Migration</td><td>a migration, crash recovery, process restart, device replacement, or restored computational snapshot is prevented from restoring a previously consumed finality ...</td><td>none named in the claim text</td></tr>
<tr><td>59</td><td>Dependent</td><td>53</td><td>Migration</td><td>a plurality of un nished Candidate Acts are transferred between computing environments and each Candidate Act is associated with a separately verifiable operation scope, ...</td><td>none named in the claim text</td></tr>
<tr><td>60</td><td>Dependent</td><td>53</td><td>Migration</td><td>the source and destination execution environments comprise respective combinations of mobile devices, virtual machines, confidential-computing environments, cloud ...</td><td>hardware terms</td></tr>
</tbody>
</table>
<t>Group-level fields (necessary components, operative condition, enforcement boundary, existing coverage, prior-art terms) follow. A per-claim chart of these fields was not produced; candidate additional limitations and per-claim implementation examples are not given, to avoid inventing limitations.</t>
<table>
<thead><tr><th>Group</th><th>Technical focus</th><th>What existing work covers</th><th>Search terms</th></tr></thead>
<tbody>
<tr><td>System (1-30)</td><td>Heterogeneous authorization translation and protected actual-effect enforcement</td><td>Authentication, OAuth-style authorization and attestation exist <xref target="RFC6749"/> <xref target="RFC9334"/>; the claim concerns translation to a canonical descriptor and sink-side reconstruction.</td><td>cross platform execution finality; permission translation; canonical effect descriptor; destination side verification</td></tr>
<tr><td>Method (31-42)</td><td>Operation-specific cross-platform execution-finality procedure</td><td>Same standards context.</td><td>method enforcing execution finality heterogeneous; scoped non-bearer enablement; protected consumption</td></tr>
<tr><td>Apparatus (43-52)</td><td>Receiving-platform protected Finality Sink</td><td>Resource gateways and policy enforcement points exist.</td><td>finality sink apparatus; actual pending operation reconstruction; sink local authority check</td></tr>
<tr><td>Migration (53-60)</td><td>Cross-device authority transfer and source fencing</td><td>Fencing tokens, leases and generation counters exist as techniques.</td><td>cross device authority migration; source fencing; ownership epoch; snapshot resurrection</td></tr>
</tbody>
</table>
</section>
<section anchor="app-c">
<name>Appendix C: Architecture Flowcharts</name>
<t>The 30 diagrams below are conceptual ASCII figures; each names its boundary. Mermaid versions are not included because a single XML file was requested.</t>
<figure anchor="fig-whole">
<name>Whole cross-platform protected architecture</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Originating user, application or AI agent                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Source operating system                                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Candidate Act (Non-Effective State)                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Source authority qualification                           |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Platform evidence and semantic translation adapter       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Protected Canonical Effect Descriptor                    |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Protected Enforcement Domain                             |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Mandatory conditions satisfied? NO: deny / hold /        |
| reconcile                                                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| YES: scoped non-bearer finality enablement               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Receiving-platform Finality Sink: reconstruct actual     |
| operation                                                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Sink-local authority and consumption check               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Actual consequence through a covered path                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Qualified outcome evidence / recovery                    |
+----------------------------------------------------------+

Boundary: Controls only effect paths the sink actually covers.]]></artwork>
</figure>
<figure anchor="fig-a2i">
<name>Android-to-iOS consequential operation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Android app builds Candidate Act                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Keystore-backed evidence of app and device               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Adapter -> canonical descriptor                          |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Protected domain verifies authority                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| iOS-side service sink reconstructs request               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Consume enablement; effect; outcome evidence             |
+----------------------------------------------------------+

Boundary: Sink on the receiving side; ordinary apps do not
          mediate other apps.]]></artwork>
</figure>
<figure anchor="fig-i2a">
<name>iOS-to-Android operation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| iOS app builds Candidate Act                             |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Keychain / Secure Enclave evidence                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Adapter -> canonical descriptor                          |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Protected domain verifies authority                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Android-side sink reconstructs request                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Consume enablement; effect; outcome evidence             |
+----------------------------------------------------------+

Boundary: No private kernel hooks are assumed.]]></artwork>
</figure>
<figure anchor="fig-w2c">
<name>Windows/Linux to cloud resource</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Windows or Linux workload                                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Workload identity and TPM/key evidence                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Descriptor of cloud API effect                           |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Protected domain: policy, revocation                     |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Cloud resource gate reconstructs API call                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Effect; outcome evidence                                 |
+----------------------------------------------------------+

Boundary: A direct cloud credential outside the gate is an
          uncovered path.]]></artwork>
</figure>
<figure anchor="fig-xlate">
<name>Cross-platform authority translation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Native permission or entitlement                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Platform adapter (qualified)                             |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Effect-specific authority claim                          |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Compare with canonical semantics                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Translation result: equivalent / narrower / reject       |
+----------------------------------------------------------+

Boundary: Permission names are not assumed equivalent.]]></artwork>
</figure>
<figure anchor="fig-desc">
<name>Canonical effect-descriptor construction</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Native operation request                                 |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Resolve effective resource, beneficiary, parameters      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Normalize (units, encoding, aliases)                     |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Reject unknown critical fields                           |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Bind to authority, policy, revocation generations        |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Descriptor digest and protected record                   |
+----------------------------------------------------------+

Boundary: A valid hash of a wrongly built descriptor proves
          nothing about meaning.]]></artwork>
</figure>
<figure anchor="fig-qual">
<name>Platform enforcement-capability qualification</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Security component and controlled resource               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Determine observed vs blocking vs consumption control    |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Check generation and trust assumptions                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| List known alternate effect routes                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Qualified / partial / unresolved / disqualified          |
+----------------------------------------------------------+

Boundary: Unavailable blocking leads to restriction or
          withholding, not silent weakening.]]></artwork>
</figure>
<figure anchor="fig-obs">
<name>Observation-only versus blocking interfaces</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Interface offers notification or logs?                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Yes: classify as observation only                        |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Interface can refuse the operation?                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Yes: blocking candidate; check coverage                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Move enforcement to a resource-side gateway if not       |
+----------------------------------------------------------+

Boundary: Product names do not decide the class.]]></artwork>
</figure>
<figure anchor="fig-local">
<name>Protected local authority and destination sink</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Local protected authority holds enablement               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Send operation-specific non-bearer enablement            |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Destination sink reconstructs actual operation           |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Check current authority and consumption                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Permit or refuse                                         |
+----------------------------------------------------------+

Boundary: The local authority is not the destination sink.]]></artwork>
</figure>
<figure anchor="fig-broker">
<name>Credential broker with downstream enforcement</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Workload requests credential use                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Broker checks operation binding                          |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Broker activates operation-bound credential              |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Downstream resource enforces scope                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Other credential copies remain uncovered                 |
+----------------------------------------------------------+

Boundary: Broker does not control other copies of credentials.]]></artwork>
</figure>
<figure anchor="fig-deleg">
<name>Cross-provider AI-agent delegation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Provider A agent holds authority                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Delegation lineage record with attenuation               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Provider B agent receives bounded scope                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Descriptor and enablement per operation                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Sink verifies lineage                                    |
+----------------------------------------------------------+

Boundary: Scope can only narrow.]]></artwork>
</figure>
<figure anchor="fig-nonbearer">
<name>Scoped non-bearer enablement</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Verified descriptor                                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Create enablement bound to operation, sink, state        |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Enablement is not a bearer token                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Sink checks binding at use                               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Used once; consumed                                      |
+----------------------------------------------------------+

Boundary: Possession alone confers no authority.]]></artwork>
</figure>
<figure anchor="fig-epoch">
<name>Authority and revocation epochs</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Authority generation N                                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Revocation event raises epoch                            |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Sink compares epoch at effect time                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Older epoch: refuse                                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Record decision                                          |
+----------------------------------------------------------+

Boundary: No global instant revocation is assumed.]]></artwork>
</figure>
<figure anchor="fig-replay">
<name>Protected replay and use consumption</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Enablement presented                                     |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Check monotonic consumption state                        |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Unconsumed: mark consumed atomically                     |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Consumed: refuse                                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Crash: unknown stays unresolved                          |
+----------------------------------------------------------+

Boundary: Consumption state must survive crashes.]]></artwork>
</figure>
<figure anchor="fig-migr">
<name>Source-to-destination migration</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Identify unfinished operation                            |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Authenticate target environment                          |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Move computational context (not authority)               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| New ownership generation; fence source                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Destination reconstructs scope and verifies pending      |
| operation                                                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Reconcile partial failure                                |
+----------------------------------------------------------+

Boundary: State movement is not authority transfer.]]></artwork>
</figure>
<figure anchor="fig-fence">
<name>Source-device fencing</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Source owner generation N                                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Fencing request or evidence                              |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Source confirms or sink rejects older generation         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Loss of contact is not proof of fencing                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Destination generation N+1 active                        |
+----------------------------------------------------------+

Boundary: Fencing is evidenced, not assumed.]]></artwork>
</figure>
<figure anchor="fig-offline">
<name>Offline authority and delayed revocation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Offline lease with bounded scope                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Clock uncertainty bounds applied                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Operate within limits                                    |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Reconnect: reconcile revocations                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Past limits: refuse                                      |
+----------------------------------------------------------+

Boundary: Offline limits are explicit.]]></artwork>
</figure>
<figure anchor="fig-cw2a">
<name>CW-2 bounded real effect</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Operation split into phases                              |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Phase 1 independently authorized                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Bounded real effect through a sink                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Qualified evidence obtained                              |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Validator evaluates evidence                             |
+----------------------------------------------------------+

Boundary: A test simulation is not a real first effect.]]></artwork>
</figure>
<figure anchor="fig-cw2b">
<name>CW-2 receipt-constituted continuation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Accepted effect receipt                                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Receipt becomes causal condition for phase 2             |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Successor sink checks receipt-dependent authority        |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Missing, stale or contradicted: no successor effect      |
+----------------------------------------------------------+

Boundary: A validator is not an unrestricted credential.]]></artwork>
</figure>
<figure anchor="fig-cw3">
<name>CW-3 detected deviation and reauthorization</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Cumulative actual effect vs original scope               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Independent security signals correlated                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Risk-adaptive suspension                                 |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Exact changed consequence presented                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Approval bound to actual effect; limited renewed         |
| authority                                                |
+----------------------------------------------------------+

Boundary: Approval does not override unrelated mandatory
          protections.]]></artwork>
</figure>
<figure anchor="fig-conf">
<name>Cross-agent conflict reservation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Agent A and B propose effects                            |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Effect footprints compared                               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Atomic reservation at sink                               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Conflict: one proceeds, one held                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Sink-bound fencing                                       |
+----------------------------------------------------------+

Boundary: Reservation is not execution authority.]]></artwork>
</figure>
<figure anchor="fig-alias">
<name>Resource alias and semantic referent</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| User sees object label                                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Resolve to actual resource identity                      |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Bind in descriptor                                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Re-resolve at effect time                                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Changed referent: refuse                                 |
+----------------------------------------------------------+

Boundary: Names do not identify effective resources.]]></artwork>
</figure>
<figure anchor="fig-toolup">
<name>Tool-runtime update and policy revalidation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Tool or runtime updated                                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Generation changes                                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Authority bound to old generation invalid                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Revalidate contracts and policy                          |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Resume under new authority only                          |
+----------------------------------------------------------+

Boundary: Updates are consequential transitions.]]></artwork>
</figure>
<figure anchor="fig-sandbox">
<name>Sandbox escape-effect boundary</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Operation inside sandbox                                 |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Crosses sandbox boundary                                 |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Boundary effect check                                    |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Separate controller decides                              |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Independent effect evidence                              |
+----------------------------------------------------------+

Boundary: A sandbox does not control an external service that
          accepts the credential.]]></artwork>
</figure>
<figure anchor="fig-multi">
<name>Multi-resource partial-commit reconciliation</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Effects across resources R1..Rn                          |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Per-sink commit results                                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Partial commit detected                                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Compensation as separate Candidate Act                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Reconcile; no atomicity promised                         |
+----------------------------------------------------------+

Boundary: Compensation needs its own authority.]]></artwork>
</figure>
<figure anchor="fig-irrev">
<name>Irreversible physical-effect hazard qualification</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Candidate irreversible command                           |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Hazard and state evidence                                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Pre-effect hazard check                                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Fail: contain and hold                                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Pass: scoped enablement; observe effect                  |
+----------------------------------------------------------+

Boundary: Not a safety certification.]]></artwork>
</figure>
<figure anchor="fig-obsrv">
<name>Compromised-observer evidence checking</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Observer reports effect                                  |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Check observer qualification and coverage                |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Independent attestation or second source                 |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Conflict: UNRESOLVED                                     |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Evidence record                                          |
+----------------------------------------------------------+

Boundary: Misleading observers are assumed possible.]]></artwork>
</figure>
<figure anchor="fig-prec">
<name>Policy precedence among several resource owners</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Several principals and owners                            |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Collect applicable policies                              |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Apply precedence rules                                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Mandatory consent present?                               |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Permit or hold                                           |
+----------------------------------------------------------+

Boundary: Unresolved conflict withholds the effect.]]></artwork>
</figure>
<figure anchor="fig-closure">
<name>Protected evidence and authorization closure</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Authorization record                                     |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Consumption record                                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Outcome evidence                                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Closure check for the selected operation                 |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Closed or open with reasons                              |
+----------------------------------------------------------+

Boundary: Closure requires all required evidence.]]></artwork>
</figure>
<figure anchor="fig-recov">
<name>Failure, rollback and crash-consistent recovery</name>
<artwork type="ascii-art"><![CDATA[+----------------------------------------------------------+
| Failure or crash                                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Read journal and consumption state                       |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Outcome known? NO: indeterminate                         |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| Recovery Candidate Act (new authority)                   |
+----------------------------------------------------------+
                              |
                              v
+----------------------------------------------------------+
| No blind replay                                          |
+----------------------------------------------------------+

Boundary: No universal rollback is promised.]]></artwork>
</figure>
</section>
<section anchor="app-d">
<name>Appendix D: Data Models and Pseudocode</name>
<t>Records are transport-neutral field lists. No mandatory universal encoding is defined; CBOR or JSON may be used where an interoperability case justifies it.</t>
<dl newline="true">
<dt>SourceAuthorityRecord</dt>
<dd><t>original authority; principal; source platform; delegated scope; current generation; restrictions</t></dd>
<dt>PlatformQualificationRecord</dt>
<dd><t>native permission; actual security capability; protected component; effect-path scope; qualification generation</t></dd>
<dt>CandidateActRecord</dt>
<dd><t>proposed operation; original task; target; effect class; initiating workload</t></dd>
<dt>CanonicalEffectDescriptor</dt>
<dd><t>material executable effect semantics (see the main text)</t></dd>
<dt>ProtectedValidationEvidence</dt>
<dd><t>qualified source; predicate; current state; protected evidence</t></dd>
<dt>ScopedFinalityEnablement</dt>
<dd><t>operation-specific, sink-bound, non-bearer permission under protected state</t></dd>
<dt>AuthorityTransferRecord</dt>
<dd><t>source owner; destination owner; epochs; fencing evidence; transfer status</t></dd>
<dt>ResourceSinkRecord</dt>
<dd><t>sink identity; controlled effect path; operation visibility; consumption state</t></dd>
<dt>PriorEffectReceipt</dt>
<dd><t>qualified evidence of a bounded earlier effect</t></dd>
<dt>ContinuationAuthorityRecord</dt>
<dd><t>successor-specific, receipt-dependent authorization (CW-2)</t></dd>
<dt>ReauthorizationRecord</dt>
<dd><t>current revised permission (CW-3)</t></dd>
<dt>OutcomeEvidenceRecord</dt>
<dd><t>source-qualified effect evidence; unresolved outcome classification</t></dd>
<dt>ClosureRecord</dt>
<dd><t>required enforcement, authorization, consumption and outcome evidence for the operation</t></dd>
</dl>
<figure anchor="alg-sink">
<name>Finality Sink decision (illustrative)</name>
<sourcecode type="pseudocode"><![CDATA[FUNCTION SINK_DECIDE(enablement, pending_op, state):
  desc = RECONSTRUCT(pending_op)
  IF desc IS NONE: RETURN REFUSE_UNRESOLVED
  IF DIGEST(desc) != enablement.descriptor_digest: RETURN REFUSE
  IF enablement.sink != SELF: RETURN REFUSE
  IF state.revocation_epoch > enablement.revocation_gen: RETURN
    REFUSE
  IF state.owner_generation != enablement.owner_gen: RETURN REFUSE
  IF NOT COVERED_PATH(pending_op): RETURN REFUSE_NOT_COVERED
  IF NOT CONSUME_ATOMIC(enablement.id): RETURN REFUSE_CONSUMED
  RETURN PERMIT]]></sourcecode>
</figure>
<figure anchor="alg-qual">
<name>Enforcement qualification (illustrative)</name>
<sourcecode type="pseudocode"><![CDATA[FUNCTION QUALIFY(component, op_class):
  caps = MEASURE_CAPABILITIES(component)
  IF NOT caps.can_block: RETURN DISQUALIFIED_FOR_BLOCKING
  IF ALTERNATE_ROUTES(component, op_class): RETURN PARTIAL
  IF NOT FRESH(component.generation): RETURN UNRESOLVED
  RETURN QUALIFIED]]></sourcecode>
</figure>
<figure anchor="alg-migr">
<name>Cross-device migration (illustrative)</name>
<sourcecode type="pseudocode"><![CDATA[FUNCTION MIGRATE(op, source, dest):
  IF NOT AUTHENTICATED(dest): RETURN ABORT
  gen = NEW_OWNER_GENERATION(op)
  IF NOT FENCED(source, gen): RETURN HOLD_UNRESOLVED
  scope = RECONSTRUCT_SCOPE(op, dest)
  enable = ISSUE_ENABLEMENT(op, dest, gen, scope)
  RETURN enable  // state copy confers no authority]]></sourcecode>
</figure>
<figure anchor="alg-desc">
<name>Descriptor construction (illustrative)</name>
<sourcecode type="pseudocode"><![CDATA[FUNCTION BUILD_DESCRIPTOR(native_op):
  r = RESOLVE_EFFECTIVE_RESOURCE(native_op)
  FOR f IN CRITICAL_FIELDS(native_op.class):
    IF f NOT IN native_op: RETURN REJECT
  d = NORMALIZE(native_op, r)
  IF HAS_UNKNOWN_CRITICAL(d): RETURN REJECT
  RETURN d]]></sourcecode>
</figure>
</section>
<section anchor="app-e">
<name>Appendix E: Threats and Negative Tests</name>
<t>Sixty threats. Embodiment references are the editor's mapping. Fields for failure response, residual risk and actual-effect observation use common wording; attacker capability, boundary, invariant, check and test are specific to each entry. Tests are ideas, not executed results.</t>
<section>
<name>1. Authenticated but unauthorized operation</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Holds valid credentials</t></dd>
<dt>Affected boundary</dt>
<dd><t>Sink</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E15</t></dd>
<dt>Protected invariant</dt>
<dd><t>Authentication is not operation authority</t></dd>
<dt>Required technical check</dt>
<dd><t>Sink reconstructs operation and checks scoped enablement</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Valid credential, no enablement: effect refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>2. Source permission incorrectly translated</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Misconfigured or malicious adapter</t></dd>
<dt>Affected boundary</dt>
<dd><t>Adapter</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E6</t></dd>
<dt>Protected invariant</dt>
<dd><t>Translation preserves meaning</t></dd>
<dt>Required technical check</dt>
<dd><t>Qualify adapter generation; compare descriptor to native evidence</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Adapter widens scope: validation fails</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>3. Overbroad destination permission</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Destination grants more than source</t></dd>
<dt>Affected boundary</dt>
<dd><t>Destination sink</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E9</t></dd>
<dt>Protected invariant</dt>
<dd><t>Destination scope &lt;= source scope</t></dd>
<dt>Required technical check</dt>
<dd><t>Reconstruct destination scope from source authority</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Broader destination right: not enabled</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>4. Native entitlement mistaken for final authority</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>App with entitlement</t></dd>
<dt>Affected boundary</dt>
<dd><t>Platform</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E15</t></dd>
<dt>Protected invariant</dt>
<dd><t>Entitlement is advice, not finality</t></dd>
<dt>Required technical check</dt>
<dd><t>Require scoped non-bearer enablement</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Entitled app without enablement: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>5. Valid credential used on wrong target</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Holds credential</t></dd>
<dt>Affected boundary</dt>
<dd><t>Resource</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E21</t></dd>
<dt>Protected invariant</dt>
<dd><t>Effective resource identity matches</t></dd>
<dt>Required technical check</dt>
<dd><t>Resolve actual resource at the sink</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Credential for A used on B: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>6. Malicious authorization adapter</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Controls an adapter</t></dd>
<dt>Affected boundary</dt>
<dd><t>Adapter</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E38</t></dd>
<dt>Protected invariant</dt>
<dd><t>Adapters are qualified, not trusted by default</t></dd>
<dt>Required technical check</dt>
<dd><t>Independent attestation of adapter generation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Unqualified adapter: UNRESOLVED</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>7. Missing critical effect fields</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Omits fields</t></dd>
<dt>Affected boundary</dt>
<dd><t>Descriptor</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E6</t></dd>
<dt>Protected invariant</dt>
<dd><t>Unknown or missing critical fields reject</t></dd>
<dt>Required technical check</dt>
<dd><t>Reject on missing critical field</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Drop beneficiary field: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>8. Semantic mismatch between platforms</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Exploits differing meaning</t></dd>
<dt>Affected boundary</dt>
<dd><t>Adapter/descriptor</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E6</t></dd>
<dt>Protected invariant</dt>
<dd><t>Equivalence is semantic, not syntactic</t></dd>
<dt>Required technical check</dt>
<dd><t>Contract-compatibility validation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Same name, different effect: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>9. Invalid resource alias</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Alias manipulation</t></dd>
<dt>Affected boundary</dt>
<dd><t>Resource</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E21</t></dd>
<dt>Protected invariant</dt>
<dd><t>Alias resolves to approved resource</t></dd>
<dt>Required technical check</dt>
<dd><t>Resolve at effectuation time</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Alias rebinding: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>10. Redirected destination</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Redirect injection</t></dd>
<dt>Affected boundary</dt>
<dd><t>Destination</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E21</t></dd>
<dt>Protected invariant</dt>
<dd><t>Final destination equals approved</t></dd>
<dt>Required technical check</dt>
<dd><t>Check effective destination</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Redirect to other host: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>11. Wrong effective principal</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Confuses principals</t></dd>
<dt>Affected boundary</dt>
<dd><t>Delegation</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E7</t></dd>
<dt>Protected invariant</dt>
<dd><t>Effective principal equals authorised</t></dd>
<dt>Required technical check</dt>
<dd><t>Bind principal in descriptor</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Delegate acts as other principal: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>12. Delegation scope expansion</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Nested agent</t></dd>
<dt>Affected boundary</dt>
<dd><t>Delegation chain</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E28</t></dd>
<dt>Protected invariant</dt>
<dd><t>Attenuation only</t></dd>
<dt>Required technical check</dt>
<dd><t>Verify lineage attenuation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Child exceeds parent: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>13. Forged agent authority</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Forger</t></dd>
<dt>Affected boundary</dt>
<dd><t>Delegation</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E7</t></dd>
<dt>Protected invariant</dt>
<dd><t>Authority has verifiable lineage</t></dd>
<dt>Required technical check</dt>
<dd><t>Verify cryptographic lineage</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Forged delegation: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>14. Agent self-reported success</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Agent</t></dd>
<dt>Affected boundary</dt>
<dd><t>Evidence</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E4</t></dd>
<dt>Protected invariant</dt>
<dd><t>Self-report is not qualified evidence</t></dd>
<dt>Required technical check</dt>
<dd><t>Require source-qualified evidence</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Self-reported success: UNRESOLVED</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>15. Untrusted tool-output instruction</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Tool output author</t></dd>
<dt>Affected boundary</dt>
<dd><t>Model-driven execution</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E24</t></dd>
<dt>Protected invariant</dt>
<dd><t>Output is data, not authority</t></dd>
<dt>Required technical check</dt>
<dd><t>Track instruction provenance</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Output instructs transfer: no enablement</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>16. Prompt injection</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Injected text</t></dd>
<dt>Affected boundary</dt>
<dd><t>Model input</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E24</t></dd>
<dt>Protected invariant</dt>
<dd><t>Instruction authority has provenance</t></dd>
<dt>Required technical check</dt>
<dd><t>Provenance check before effect</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Injected instruction cannot enable</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>17. Tool-runtime substitution</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Swaps runtime</t></dd>
<dt>Affected boundary</dt>
<dd><t>Tool lifecycle</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E26</t></dd>
<dt>Protected invariant</dt>
<dd><t>Runtime generation continuity</t></dd>
<dt>Required technical check</dt>
<dd><t>Revalidate on generation change</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Swapped runtime: revalidate</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>18. Changed executable generation</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Updates tool</t></dd>
<dt>Affected boundary</dt>
<dd><t>Runtime</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E10</t></dd>
<dt>Protected invariant</dt>
<dd><t>Authority bound to generation</t></dd>
<dt>Required technical check</dt>
<dd><t>Effectuation-time generation check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Old authority on new generation: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>19. Stale model or algorithm fingerprint</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Model drift</t></dd>
<dt>Affected boundary</dt>
<dd><t>Runtime</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E10</t></dd>
<dt>Protected invariant</dt>
<dd><t>Fingerprint current at effect time</t></dd>
<dt>Required technical check</dt>
<dd><t>Compare fingerprint at sink</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Stale fingerprint: revalidate</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>20. Incomplete enforcement coverage</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Uses uncovered path</t></dd>
<dt>Affected boundary</dt>
<dd><t>Coverage</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E14</t></dd>
<dt>Protected invariant</dt>
<dd><t>Claims limited to covered paths</t></dd>
<dt>Required technical check</dt>
<dd><t>Coverage controller reports paths</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Uncovered path: claim not made</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>21. Observation-only interface misclassified as blocking</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Misclassification</t></dd>
<dt>Affected boundary</dt>
<dd><t>Qualification</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E14</t></dd>
<dt>Protected invariant</dt>
<dd><t>Capability classes differ</t></dd>
<dt>Required technical check</dt>
<dd><t>Qualification record</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Monitor API marked blocking: disqualify</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>22. Alternate API bypass</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Uses another API</t></dd>
<dt>Affected boundary</dt>
<dd><t>Effect path</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E14</t></dd>
<dt>Protected invariant</dt>
<dd><t>All effect paths covered or disclosed</t></dd>
<dt>Required technical check</dt>
<dd><t>Path discovery and coverage</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Alternate API used: detected or disclosed</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>23. Exported secret used outside broker</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Exports secret</t></dd>
<dt>Affected boundary</dt>
<dd><t>Credential custody</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E23</t></dd>
<dt>Protected invariant</dt>
<dd><t>Non-exportable custody</t></dd>
<dt>Required technical check</dt>
<dd><t>Broker-bound activation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Exported key: not accepted by sink</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>24. Stale source device after migration</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Old device</t></dd>
<dt>Affected boundary</dt>
<dd><t>Source</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E9</t></dd>
<dt>Protected invariant</dt>
<dd><t>Source fenced</t></dd>
<dt>Required technical check</dt>
<dd><t>Fence and generation check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Old device acts: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>25. Snapshot authority resurrection</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Restores snapshot</t></dd>
<dt>Affected boundary</dt>
<dd><t>Migration</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E27</t></dd>
<dt>Protected invariant</dt>
<dd><t>State copy is not authority</t></dd>
<dt>Required technical check</dt>
<dd><t>Generation fencing</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Restored worker refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>26. Duplicate cross-device execution</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Two devices</t></dd>
<dt>Affected boundary</dt>
<dd><t>Sink</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E9</t></dd>
<dt>Protected invariant</dt>
<dd><t>Single owner generation</t></dd>
<dt>Required technical check</dt>
<dd><t>Sink consumption check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Second execution refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>27. Unverified source fencing</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Claims fenced</t></dd>
<dt>Affected boundary</dt>
<dd><t>Source</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E9</t></dd>
<dt>Protected invariant</dt>
<dd><t>Fencing is evidenced</t></dd>
<dt>Required technical check</dt>
<dd><t>Require fencing evidence</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>No evidence: UNRESOLVED</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>28. Destination-side weaker enforcement</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Weak destination</t></dd>
<dt>Affected boundary</dt>
<dd><t>Destination</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E9</t></dd>
<dt>Protected invariant</dt>
<dd><t>Enforcement capability equivalent or restricted</t></dd>
<dt>Required technical check</dt>
<dd><t>Qualification comparison</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Weaker destination: restrict or withhold</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>29. Cross-domain token replay</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Replays token</t></dd>
<dt>Affected boundary</dt>
<dd><t>Sink</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E15</t></dd>
<dt>Protected invariant</dt>
<dd><t>Non-bearer enablement</t></dd>
<dt>Required technical check</dt>
<dd><t>Bind to operation and sink</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Replay elsewhere: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>30. Consumed authorization replay</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Replays consumed</t></dd>
<dt>Affected boundary</dt>
<dd><t>Sink</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E16</t></dd>
<dt>Protected invariant</dt>
<dd><t>One-time consumption</t></dd>
<dt>Required technical check</dt>
<dd><t>Monotonic consumption</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Second use: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>31. Revocation epoch rollback</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Rolls back state</t></dd>
<dt>Affected boundary</dt>
<dd><t>Revocation</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E18</t></dd>
<dt>Protected invariant</dt>
<dd><t>Epochs monotonic</t></dd>
<dt>Required technical check</dt>
<dd><t>Verify epoch monotonicity</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Older epoch: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>32. Cached permission after revocation</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Uses cache</t></dd>
<dt>Affected boundary</dt>
<dd><t>Sink</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E18</t></dd>
<dt>Protected invariant</dt>
<dd><t>Revocation checked at effect</t></dd>
<dt>Required technical check</dt>
<dd><t>Check current revocation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Cached allow after revoke: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>33. Offline stale authority</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Disconnected device</t></dd>
<dt>Affected boundary</dt>
<dd><t>Offline</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E35</t></dd>
<dt>Protected invariant</dt>
<dd><t>Offline limits explicit</t></dd>
<dt>Required technical check</dt>
<dd><t>Bounded offline authority</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Past limit: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>34. Incorrect lease expiry</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Clock error</t></dd>
<dt>Affected boundary</dt>
<dd><t>Lease</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E17</t></dd>
<dt>Protected invariant</dt>
<dd><t>Lease honours clock uncertainty</t></dd>
<dt>Required technical check</dt>
<dd><t>Use uncertainty bounds</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Near expiry: conservative</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>35. Clock manipulation</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Alters clock</t></dd>
<dt>Affected boundary</dt>
<dd><t>Time</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E17</t></dd>
<dt>Protected invariant</dt>
<dd><t>Time not solely trusted</t></dd>
<dt>Required technical check</dt>
<dd><t>Independent time or counters</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Skewed clock: reject</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>36. Conflicting resource owners</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Competing owners</t></dd>
<dt>Affected boundary</dt>
<dd><t>Resource</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E30</t></dd>
<dt>Protected invariant</dt>
<dd><t>Precedence resolved</t></dd>
<dt>Required technical check</dt>
<dd><t>Owner consent check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Conflicting owners: UNRESOLVED</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>37. Missing mandatory owner consent</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Omits owner</t></dd>
<dt>Affected boundary</dt>
<dd><t>Resource</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E39</t></dd>
<dt>Protected invariant</dt>
<dd><t>Mandatory consent present</t></dd>
<dt>Required technical check</dt>
<dd><t>Consent check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>No consent: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>38. Resource quota oversubscription</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Parallel requests</t></dd>
<dt>Affected boundary</dt>
<dd><t>Budget</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E22</t></dd>
<dt>Protected invariant</dt>
<dd><t>Cumulative limits</t></dd>
<dt>Required technical check</dt>
<dd><t>Atomic reservation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Over budget: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>39. Concurrent effect conflict</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Concurrent actors</t></dd>
<dt>Affected boundary</dt>
<dd><t>Sink</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E11</t></dd>
<dt>Protected invariant</dt>
<dd><t>Version-bound reservations</t></dd>
<dt>Required technical check</dt>
<dd><t>Sink commit validation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Conflict: one refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>40. Unauthorized tool-chain continuation</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Skips step</t></dd>
<dt>Affected boundary</dt>
<dd><t>Workflow</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E20</t></dd>
<dt>Protected invariant</dt>
<dd><t>Dependencies satisfied</t></dd>
<dt>Required technical check</dt>
<dd><t>Dependency-bound authority</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Skipped step: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>41. Missing prior-effect receipt</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Omits receipt</t></dd>
<dt>Affected boundary</dt>
<dd><t>CW-2</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E2</t></dd>
<dt>Protected invariant</dt>
<dd><t>Receipt required for continuation</t></dd>
<dt>Required technical check</dt>
<dd><t>Successor sink checks receipt</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>No receipt: no successor effect</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>42. Replayed Interim Effectuation Validator receipt</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Replays</t></dd>
<dt>Affected boundary</dt>
<dd><t>CW-2</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E2</t></dd>
<dt>Protected invariant</dt>
<dd><t>Receipt bound to phase and operation</t></dd>
<dt>Required technical check</dt>
<dd><t>Bind receipt context</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Replayed receipt: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>43. Wrong validator identity</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Impersonates validator</t></dd>
<dt>Affected boundary</dt>
<dd><t>CW-2</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E2</t></dd>
<dt>Protected invariant</dt>
<dd><t>Validator identity authenticated</t></dd>
<dt>Required technical check</dt>
<dd><t>Identity check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Wrong validator: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>44. Untrusted human approval interface</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Spoofed UI</t></dd>
<dt>Affected boundary</dt>
<dd><t>Approval</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E19</t></dd>
<dt>Protected invariant</dt>
<dd><t>Approval bound to actual operation</t></dd>
<dt>Required technical check</dt>
<dd><t>Authenticated interface state</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Spoofed UI: not accepted</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>45. Changed actual operation after approval</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Mutates request</t></dd>
<dt>Affected boundary</dt>
<dd><t>Approval</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E19</t></dd>
<dt>Protected invariant</dt>
<dd><t>Approved equals executed</t></dd>
<dt>Required technical check</dt>
<dd><t>Sink reconstruction</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Post-approval change: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>46. Credential rotation race</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Race</t></dd>
<dt>Affected boundary</dt>
<dd><t>Credential</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E31</t></dd>
<dt>Protected invariant</dt>
<dd><t>Generation fencing</t></dd>
<dt>Required technical check</dt>
<dd><t>Generation check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Old generation: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>47. Secret-custody compromise</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Compromises custody</t></dd>
<dt>Affected boundary</dt>
<dd><t>Custody</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E23</t></dd>
<dt>Protected invariant</dt>
<dd><t>Custody qualification</t></dd>
<dt>Required technical check</dt>
<dd><t>Attest custody</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Unqualified custody: restrict</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>48. Policy update self-authorization</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Agent edits policy</t></dd>
<dt>Affected boundary</dt>
<dd><t>Policy</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E33</t></dd>
<dt>Protected invariant</dt>
<dd><t>Mutations not self-authorizing</t></dd>
<dt>Required technical check</dt>
<dd><t>Independent approval of policy change</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Self-approved change: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>49. Sandbox-to-host privilege transition</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Escapes sandbox</t></dd>
<dt>Affected boundary</dt>
<dd><t>Sandbox</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E25</t></dd>
<dt>Protected invariant</dt>
<dd><t>Effects crossing boundary checked</t></dd>
<dt>Required technical check</dt>
<dd><t>Boundary effect check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Escaping op: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>50. Hardware command bypass</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Direct device access</t></dd>
<dt>Affected boundary</dt>
<dd><t>Hardware</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E14</t></dd>
<dt>Protected invariant</dt>
<dd><t>Command path coverage</t></dd>
<dt>Required technical check</dt>
<dd><t>Hardware gate or disclose gap</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Direct command: refused or disclosed</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>51. Device reset and queue replay</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Resets device</t></dd>
<dt>Affected boundary</dt>
<dd><t>Hardware</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E16</t></dd>
<dt>Protected invariant</dt>
<dd><t>No replay after reset</t></dd>
<dt>Required technical check</dt>
<dd><t>Persistent consumption</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Replayed queue: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>52. Privacy-budget overflow</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Many small disclosures</t></dd>
<dt>Affected boundary</dt>
<dd><t>Privacy</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E8</t></dd>
<dt>Protected invariant</dt>
<dd><t>Cumulative exposure</t></dd>
<dt>Required technical check</dt>
<dd><t>Compositional budget</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Aggregate over limit: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>53. Misleading observer evidence</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Observer lies</t></dd>
<dt>Affected boundary</dt>
<dd><t>Evidence</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E38</t></dd>
<dt>Protected invariant</dt>
<dd><t>Observer qualification</t></dd>
<dt>Required technical check</dt>
<dd><t>Independent attestation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Misleading evidence: UNRESOLVED</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>54. Forged remote-effect receipt</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Forger</t></dd>
<dt>Affected boundary</dt>
<dd><t>Evidence</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E4</t></dd>
<dt>Protected invariant</dt>
<dd><t>Receipt from source-authoritative signer</t></dd>
<dt>Required technical check</dt>
<dd><t>Verify signer capability</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Receipt outside signer observation: not accepted</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>55. Lost acknowledgment</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Network loss</t></dd>
<dt>Affected boundary</dt>
<dd><t>Recovery</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E16</t></dd>
<dt>Protected invariant</dt>
<dd><t>Unknown outcome stays unknown</t></dd>
<dt>Required technical check</dt>
<dd><t>Indeterminate state</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>No ack: UNRESOLVED, no blind retry</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>56. Duplicate recovery execution</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Repeated recovery</t></dd>
<dt>Affected boundary</dt>
<dd><t>Recovery</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E12</t></dd>
<dt>Protected invariant</dt>
<dd><t>Recovery authority non-replayable</t></dd>
<dt>Required technical check</dt>
<dd><t>Consume recovery authority</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Second recovery: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>57. Partial multi-resource effect</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Partial failure</t></dd>
<dt>Affected boundary</dt>
<dd><t>Transaction</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E36</t></dd>
<dt>Protected invariant</dt>
<dd><t>Partial state contained</t></dd>
<dt>Required technical check</dt>
<dd><t>Cross-sink reconciliation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Partial commit: contained</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>58. Unauthorized compensation</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Rolls back without authority</t></dd>
<dt>Affected boundary</dt>
<dd><t>Recovery</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E12</t></dd>
<dt>Protected invariant</dt>
<dd><t>Compensation separately authorised</t></dd>
<dt>Required technical check</dt>
<dd><t>Compensation Candidate Act</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Unauthorised rollback: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>59. Irreversible unsafe operation</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Triggers irreversible act</t></dd>
<dt>Affected boundary</dt>
<dd><t>Physical</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E37</t></dd>
<dt>Protected invariant</dt>
<dd><t>Hazard check before effect</t></dd>
<dt>Required technical check</dt>
<dd><t>Consequence-bound hazard validation</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Unsafe irreversible: refused</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
<section>
<name>60. Unverified formal evidence closure</name>
<dl newline="true">
<dt>Attacker capability</dt>
<dd><t>Claims closure</t></dd>
<dt>Affected boundary</dt>
<dd><t>Closure</t></dd>
<dt>Relevant embodiment</dt>
<dd><t>E40</t></dd>
<dt>Protected invariant</dt>
<dd><t>Closure requires evidence</t></dd>
<dt>Required technical check</dt>
<dd><t>Closure record check</t></dd>
<dt>Failure response</dt>
<dd><t>Hold the operation in the Non-Effective State, reject or mark UNRESOLVED, and preserve the evidence; do not fall back to weaker protection.</t></dd>
<dt>Residual risk</dt>
<dd><t>Cannot cover effect paths that are not under control, or evidence sources that never expose the fact.</t></dd>
<dt>Concrete test</dt>
<dd><t>Missing evidence: not closed</t></dd>
<dt>Actual-effect observation requirement</dt>
<dd><t>Observe the actual effect or its absence at the covered boundary; a missing observation is recorded as unknown, not as success.</t></dd>
</dl>
</section>
</section>
<section anchor="app-f">
<name>Appendix F: Examiner Keywords and Search Phrases</name>
<t>Non-normative. Terms were composed by the editor from the claims, embodiment subjects and the source's own term lists, with industry synonyms added. Clusters map to embodiment and claim ranges at cluster level, not term by term.</t>
<t>Keywords (214):</t>
<table>
<thead><tr><th>Cluster</th><th>Embodiments / claims</th><th>Keywords</th></tr></thead>
<tbody>
<tr><td>Core architecture</td><td>1-60</td><td>Protected Execution Finality, Candidate Act, Actual Pending Operation, Non-Effective State, Protected Enforcement Domain, Finality Gate, Finality Sink, Scoped Finality Authority, Non-Bearer Execution Handle, Consequence-Bound Authorization, Source-Authority Envelope, Actual-Effect Reconstruction, execution finality, policy enforcement point, resource-side enforcement, operation reconstruction, authorization-to-effect translation</td></tr>
<tr><td>Cross-platform interoperability</td><td>1, 31, 43</td><td>Cross-Platform Execution Authority, Android iOS Interoperability, Heterogeneous Authorization Semantics, Permission Translation, Canonical Effect Descriptor, Native Security Substrate, Cross-Platform Effect Equivalence, Destination-Side Enforcement, Receiving-Platform Verification, heterogeneous operating systems, mobile desktop cloud embedded, vendor-independent security semantics, cross-OS authorization, permission mapping, entitlement mapping, capability translation</td></tr>
<tr><td>Platform assurance</td><td>1-52</td><td>Enforcement-Capability Qualification, Mandatory Blocking Capability, Observation-Only Security Interface, Effectuation-Path Coverage, Alternative Effect Path, Credential Broker, Protected Application Dispatcher, Source Security Evidence, Protected Service Identity, policy decision point, security API capability, coverage gap, bypass path, mandatory access control, hardware-backed keystore, StrongBox, Secure Enclave, TPM, HSM, KMS</td></tr>
<tr><td>Autonomous agents</td><td>E7, E20, E24, E28</td><td>Cross-Provider Agent Delegation, Agent Tool Execution, MCP Security, Agent-to-Agent Authorization, Multi-Agent Orchestration, Tool Semantic Continuity, Agent Handoff, AI Runtime Security, Agent Context Migration, Nested Tool Authority, agent sandbox, tool gateway, prompt injection, tool-output provenance, agent SDK, A2A, Model Context Protocol, delegation chain, transitive attenuation, agent identity</td></tr>
<tr><td>Cryptographic authorization</td><td>E15, E23</td><td>Sender-Constrained Capability, Scope-Limited Authorization, Protected Consumption, Revocation Epoch, Authority Generation, Replay-Resistant Permission, Trusted Execution Environment, Hardware-Rooted Authority, Attestation, proof of possession, DPoP, token exchange, rich authorization request, CBOR, COSE, JWT, CWT, mutual TLS, workload identity, capability token, one-time authorization</td></tr>
<tr><td>Resource and identity continuity</td><td>E21, E34, E9</td><td>Resource Alias Binding, Dynamic Namespace Rebinding, Object Identity, Redirect Verification, Original Source Principal, Resource Ownership Epoch, Cross-Device Authority Migration, Stale Source Fencing, Snapshot Resurrection, effective resource identity, URL normalization, service discovery identity, alias resolution, namespace rebinding, ownership generation, fencing token, lease, generation counter, device handoff, session continuity</td></tr>
<tr><td>Evidence and outcome</td><td>E2, E4, E16, E38</td><td>Authenticated Effect Evidence, Interim Effectuation Validator, Evidence-Carrying Receipt, Receipt-Gated Continuation, Actual Effect Witness, Remote Outcome Reconciliation, Crash-Consistent Effect Journal, Unknown Outcome, Effect Provenance, indeterminate outcome, idempotency key, transactional outbox, write-ahead journal, receipt validation, observer attestation, compromised observer, outcome evidence, audit log, saga, compensation</td></tr>
<tr><td>Privacy and safety</td><td>E8, E19, E37</td><td>Cumulative Privacy Exposure, Compositional Disclosure Budget, Physical Hazard Validation, Irreversible Effect, Secure Human Approval, Transaction-Bound Consent, Protected Recovery Authority, Non-Compensable Consequence, data egress control, privacy budget, differential disclosure, secure display, what-you-see-is-what-you-sign, human-in-the-loop approval, consent binding, hazard check, safety interlock, physical actuator, recipient change, selective disclosure</td></tr>
<tr><td>Formal assurance</td><td>E40</td><td>Enforcement Composition Graph, Protected Authority-Consumption Invariant, Effect-Path Reachability, Cross-Domain Safety Invariant, Machine-Checkable Finality, Formal State Transition, Negative Effect Verification, model checking, TLA+, invariant, reachability analysis, state machine verification, closure condition, proof obligation, safety property, liveness property, refinement, temporal logic, finality closure, coverage proof</td></tr>
<tr><td>Distributed and concurrency</td><td> E1, E11, E13, E22, E29</td><td>effect conflict arbitration, atomic reservation, version-bound reservation, quorum, Byzantine agreement, consensus, approval quorum, resource quota, cumulative budget, delegated quota, sink-local commit, optimistic concurrency, compare-and-swap, two-phase commit, distributed lock, lease expiry, clock uncertainty, delayed revocation, offline operation, network partition</td></tr>
<tr><td>Runtime and lifecycle</td><td>E10, E14, E25, E26, E33</td><td>model drift, runtime generation, algorithm fingerprint, tool update, installation attestation, supply-chain integrity, sandbox escape, privilege transition, container boundary, policy evolution, configuration drift, change-impact analysis, non-self-authorizing mutation, effectuation-path discovery, bypass detection, runtime revalidation, software bill of materials, secure boot, measured boot, remote attestation</td></tr>
</tbody>
</table>
<t>Compound phrases (110):</t>
<table>
<thead><tr><th>Cluster</th><th>Phrases</th></tr></thead>
<tbody>
<tr><td>Core</td><td>protected execution finality; candidate act in non-effective state; actual pending operation reconstruction; scoped non-bearer finality enablement; finality sink verification; protected enforcement domain; consequence-bound authorization; authorization-to-effect translation; source authority envelope; actual effect reconstruction</td></tr>
<tr><td>Cross-platform</td><td>cross-platform execution authority; Android iOS effect authorization; heterogeneous authorization semantics; permission translation adapter; canonical effect descriptor; native security substrate; destination-side finality verification; receiving-platform sink; cross-platform effect equivalence; vendor-independent authority semantics</td></tr>
<tr><td>Platform</td><td>enforcement capability qualification; mandatory blocking capability; observation-only security interface; effectuation path coverage; alternate effect path; credential broker downstream enforcement; protected application dispatcher; source security evidence; platform qualification record; unavailable enforcement non-effective</td></tr>
<tr><td>Agents</td><td>cross-provider agent delegation; agent tool execution authorization; MCP protected operation semantics; agent-to-agent authority; tool contract evolution; tool semantic continuity; agent handoff authority; nested tool authority attenuation; model-driven tool output provenance; agent sandbox privilege transition</td></tr>
<tr><td>Cryptographic</td><td>sender-constrained capability; scope-limited authorization; protected authority consumption; revocation epoch check; authority generation counter; replay-resistant operation permission; hardware-rooted finality; non-exportable credential custody; one-time execution authorization; operation-bound broker activation</td></tr>
<tr><td>Identity</td><td>resource alias binding; dynamic namespace rebinding; effective resource identity; redirect destination verification; original source principal; resource ownership epoch; cross-device authority migration; stale source fencing; snapshot authority resurrection; non-resurrectable in-flight effect</td></tr>
<tr><td>Evidence</td><td>authenticated effect evidence; interim effectuation validator; receipt-gated continuation; evidence-carrying receipt; actual effect witness; remote outcome reconciliation; crash-consistent finality journal; indeterminate outcome handling; effect provenance chain; compromised observer containment</td></tr>
<tr><td>Privacy and safety</td><td>cumulative privacy exposure; compositional disclosure budget; physical hazard validation; irreversible effect containment; transaction-bound human approval; protected recovery authority; non-compensable consequence; authenticated presentation state; secure approval interface; scope-limited disclosure</td></tr>
<tr><td>Formal</td><td>enforcement composition graph; authority consumption invariant; effect path reachability; cross-domain safety invariant; machine-checkable finality closure; negative effect verification; formal state transition model; closure record verification; coverage revalidation; bypass detection</td></tr>
<tr><td>Distributed</td><td>cross-agent effect conflict arbitration; atomic effect reservation; sink-bound fencing; version-bound reservation; independent approval quorum; consensus-to-effect separation; multi-resource budget; delegated quota; partial commit containment; independently authorized compensation</td></tr>
<tr><td>Runtime</td><td>model algorithm drift revalidation; tool runtime lifecycle authority; installation-time authority attestation; policy mutation non-self-authorizing; configuration change impact; credential rotation generation fencing; offline effect authority; delayed revocation propagation; conflicting policy precedence; resource owner change revocation race</td></tr>
</tbody>
</table>
<t>Boolean search phrases (25):</t>
<ol>
<li>"cross platform" AND "execution finality"</li>
<li>"Android" AND "iOS" AND "effect authorization"</li>
<li>"heterogeneous operating systems" AND "authority continuity"</li>
<li>"permission translation" AND "actual operation"</li>
<li>"canonical effect descriptor" AND "enforcement"</li>
<li>"native permission" AND "consequence bound"</li>
<li>"source authority" AND "receiving platform"</li>
<li>"destination side" AND "finality verification"</li>
<li>"cross device migration" AND "source fencing"</li>
<li>"execution ownership epoch" AND "revocation"</li>
<li>"non bearer capability" AND "effectuation"</li>
<li>"effect path coverage" AND "alternative credential"</li>
<li>"agent tool" AND "execution time authorization"</li>
<li>"MCP" AND "protected operation semantics"</li>
<li>"tool contract evolution" AND "authority"</li>
<li>"credential broker" AND "actual effect"</li>
<li>"agent delegation" AND "non transferable authority"</li>
<li>"receipt gated continuation" AND "actual effect"</li>
<li>"resource alias" AND "authorization correspondence"</li>
<li>"human approval" AND "transaction bound"</li>
<li>"runtime update" AND "finality authority"</li>
<li>"sandbox" AND "escaping operation"</li>
<li>"offline authority" AND "delayed revocation"</li>
<li>"policy mutation" AND "non self authorizing"</li>
<li>"cross domain evidence" AND "authority consumption"</li>
</ol>
<t>Examiner discussion, to be tested against the filed texts and not decided here: the objective technical problem (keeping protected authority meaningful for an actual effect across platforms); motivation to combine the established mechanisms of the prior-art table; reasonable expectation of success given the platform limits described; functional interdependence of descriptor, qualification, sink reconstruction, consumption and fencing; and avoidance of hindsight. Isolated well-known mechanisms are not claimed as novel.</t>
</section>
<section anchor="app-g">
<name>Appendix G: Industry and Product Mapping</name>
<t>Prospective, non-normative. The rows describe where a check of this kind would sit if integrated, and what it would examine. Product capabilities were not verified against current official documentation in this session, and no adoption, endorsement, implementation, copying or infringement is asserted. The alignment is AI-generated and must be cross-checked against the original PCT filings. Latency and legacy compatibility depend on the loop tier and on whether an existing resource-side gate exists; required changes are the addition of operation reconstruction and consumption at the sink.</t>
<section>
<name>Apple (iOS, iPadOS, macOS)</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>App-level dispatcher with Secure Enclave-held keys, and a server-side resource sink</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Operation-specific enablement tied to the reconstructed request</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Calls using app-held credentials outside the dispatcher</t></dd>
<dt>Why it matters</dt>
<dd><t>Platform identity of an app does not name the exact effect</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>No system-wide interception by ordinary apps</t></dd>
</dl>
</section>
<section>
<name>Google Android</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>App dispatcher with Keystore/StrongBox keys; enterprise-managed gateways; server sink</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Same, plus device-state evidence</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Unmanaged routes and exported credentials</t></dd>
<dt>Why it matters</dt>
<dd><t>Different OEM capabilities need qualification records</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Ordinary apps cannot mediate other apps</t></dd>
</dl>
</section>
<section>
<name>Microsoft Windows and Azure</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Protected service broker with TPM-backed keys; Azure resource-side checks</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Operation binding and revocation generation</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Principals that can export credentials</t></dd>
<dt>Why it matters</dt>
<dd><t>Identity differs from operation-specific authority</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Local administrators outside the broker</t></dd>
</dl>
</section>
<section>
<name>Linux ecosystems</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Resource controller with workload identity and LSM/seccomp confinement</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Operation reconstruction at the controller</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Direct credentials held by workloads</t></dd>
<dt>Why it matters</dt>
<dd><t>Process confinement does not know the effect</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Fully compromised kernel</t></dd>
</dl>
</section>
<section>
<name>Samsung, Qualcomm, MediaTek</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Hardware-backed key stores and TEEs as enforcement substrate</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Attestable enforcement generation</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Normal-world paths</t></dd>
<dt>Why it matters</dt>
<dd><t>Capability differs by chipset and OEM</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Chipset firmware behaviour</t></dd>
</dl>
</section>
<section>
<name>OpenAI Agents SDK</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Tool-call dispatcher gate before executing a tool with consequences</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Canonical effect descriptor and enablement per call</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Direct API keys held by the tool</t></dd>
<dt>Why it matters</dt>
<dd><t>Tool names do not identify effects</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>What a hosted tool does internally</t></dd>
</dl>
</section>
<section>
<name>Anthropic Claude Agent SDK and MCP</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Tool-dispatch layer and MCP server resource side</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Operation provenance and tool contract version</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Credentials available outside the dispatcher</t></dd>
<dt>Why it matters</dt>
<dd><t>Tool-output text is data, not authority</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Server-side behaviour not exposed</t></dd>
</dl>
</section>
<section>
<name>Google ADK and A2A</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Delegation hand-off between agents</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Delegation lineage with attenuation</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Direct calls between agents outside the lineage</t></dd>
<dt>Why it matters</dt>
<dd><t>Cross-provider handoff must not widen scope</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Remote agent internal state</t></dd>
</dl>
</section>
<section>
<name>Microsoft agent services</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Agent runtime dispatcher with managed identity</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Runtime-generation revalidation</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Managed identity used elsewhere</t></dd>
<dt>Why it matters</dt>
<dd><t>Identity is not operation authority</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Service internals</t></dd>
</dl>
</section>
<section>
<name>AWS agent systems and AWS cloud</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Resource-side gate beside IAM; KMS-held keys</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Sink reconstruction of the API call</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Paths using unrestricted keys</t></dd>
<dt>Why it matters</dt>
<dd><t>IAM authorizes principals, not specific effect</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Services not routed through the gate</t></dd>
</dl>
</section>
<section>
<name>Salesforce, IBM watsonx, Alibaba AgentScope</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Agent action dispatcher</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Action-contract compatibility</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Direct connectors</t></dd>
<dt>Why it matters</dt>
<dd><t>Different tool contracts under similar names</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Connector internals</t></dd>
</dl>
</section>
<section>
<name>Oracle, IBM Cloud and other clouds</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Resource gateway and HSM/KMS custody</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Consumed authorization and revocation epochs</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Other credential routes</t></dd>
<dt>Why it matters</dt>
<dd><t>Differing IAM semantics</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Provider internals</t></dd>
</dl>
</section>
<section>
<name>Arm, Intel, AMD, NVIDIA</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>TEE or confidential-computing attestation as enforcement evidence</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Attested enforcement generation</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Effects after leaving the enclave</t></dd>
<dt>Why it matters</dt>
<dd><t>Attestation shows environment, not each effect</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Microarchitectural limits</t></dd>
</dl>
</section>
<section>
<name>SAP, ServiceNow, Databricks, Snowflake</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Enterprise workflow or data-platform gate</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Dependency-bound workflow authority; disclosure budgets</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Direct database credentials</t></dd>
<dt>Why it matters</dt>
<dd><t>Data egress paths vary</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Platform-internal jobs</t></dd>
</dl>
</section>
<section>
<name>Visa, Mastercard, Swift, payment processors</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Payment execution gate beside the processor</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Reconstruction of amount, payee and currency; consumed once</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Alternate payment rails</t></dd>
<dt>Why it matters</dt>
<dd><t>Authorization of a payment is separate</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Settlement finality outside the gate</t></dd>
</dl>
</section>
<section>
<name>Huawei, Ericsson, Nokia, operators and Open RAN</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Network-management and exposure gates</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Operation reconstruction and revocation propagation</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Vendor-private interfaces</t></dd>
<dt>Why it matters</dt>
<dd><t>Multi-vendor semantics differ</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Vendor-internal behaviour</t></dd>
</dl>
</section>
<section>
<name>Siemens, Schneider Electric, industrial IoT and robotics</name>
<dl newline="true">
<dt>Proposed integration boundary</dt>
<dd><t>Actuator or command-controller gate</t></dd>
<dt>What is checked or rejected</dt>
<dd><t>Hazard check and irreversible-effect containment</t></dd>
<dt>Residual bypass risk</dt>
<dd><t>Direct actuator wiring</t></dd>
<dt>Why it matters</dt>
<dd><t>Not a safety certification</t></dd>
<dt>Unknown or not verifiable</dt>
<dd><t>Physical effects without sensors</t></dd>
</dl>
</section>
</section>
<section anchor="app-h">
<name>Appendix H: Mothership Comparison</name>
<t>This table is the editor's preliminary sorting based on the work's own description of the earlier programme. The foundational JSON (Technical_Disclosure_1.json) was not available and the Datatracker drafts were not read, so no comparison of actual filed texts was done and no patentability or priority conclusion is made.</t>
<table>
<thead><tr><th>Concept</th><th>Relationship to earlier work</th><th>Status</th></tr></thead>
<tbody>
<tr><td>Candidate Acts, Non-Effective State, Protected Enforcement Domains, Finality Sinks, scoped effect authority, revocation, anti-replay, actual-effect evidence</td><td>Foundational (Mothership programme)</td><td>Not new in October</td></tr>
<tr><td>Native-permission semantic differences</td><td>Applies earlier principles to platforms</td><td>To examine against filed texts</td></tr>
<tr><td>Platform-specific authority adapters</td><td>Specific implementation of authorization-to-effect translation</td><td>To examine</td></tr>
<tr><td>Canonical effect descriptors</td><td>Related to operation reconstruction</td><td>Cross-platform canonicalization issues may be additional</td></tr>
<tr><td>Native enforcement capability qualification</td><td>Potentially additional</td><td>To examine</td></tr>
<tr><td>Cross-platform source-authority continuity</td><td>Potentially additional</td><td>To examine</td></tr>
<tr><td>Protected cross-device authority ownership and stale-source fencing</td><td>Potentially additional</td><td>To examine</td></tr>
<tr><td>Unavailable enforcement capabilities; credential-broker limitations</td><td>Honest limitation handling</td><td>To examine</td></tr>
<tr><td>Platform upgrade and security API changes</td><td>Related to runtime lifecycle</td><td>To examine</td></tr>
<tr><td>Cross-provider AI-agent handoffs</td><td>Application of delegation continuity</td><td>To examine</td></tr>
<tr><td>Hardware and software enforcement equivalence; partial coverage and alternative paths</td><td>Potentially additional</td><td>To examine</td></tr>
</tbody>
</table>
</section>
<section anchor="app-i">
<name>Appendix I: IPR and Research Lineage</name>
<ul>
<li>Foundational: PCT/IB2026/055615, filed 4 June 2026, published 16 July 2026 as WO 2026/150382 A2 (<xref target="PCT-055615"/>; read at <xref target="TDC-12542"/>).</li>
<li>This source: Indian provisional identified by the applicant as 202631127058, reported filed 25 October 2026 (<xref target="IN-127058"/>); not verified against an official acknowledgment.</li>
<li>Earlier applicant-reported Indian provisional references: 202631129178 (29 October 2026), 202631130246 (31 October 2026), 202631131650 (3 November 2026) and 202631136555 (15 November 2026).</li>
<li>Related drafts: draft-das-execution-finality-foundations, draft-das-agent-execution-finality, draft-das-receipt-gated-iev, draft-das-consequence-bound-execution-authority (status not checked).</li>
<li>Whether later claims benefit from the PCT priority is a legal question and not decided here.</li>
</ul>
</section>
<section anchor="app-j">
<name>Appendix J: Source Verification and Limitations</name>
<section>
<name>Source Verification</name>
<ul>
<li>Source: 3,577-page PDF; text extracted with pdftotext; embodiment boundaries located from END OF EMBODIMENT markers; claims 1-60 parsed in sequence.</li>
<li>Duplicate E31: two full sections labelled Embodiment 31 (about pages 2567-2650 and 2650-2734). A sentence-level comparison found the same number of sentences in each, and the differences seen were whitespace and bullet-glyph artefacts of the extraction. They are treated as one embodiment.</li>
<li>CW-1 begins near page 15, CW-2 near page 63, CW-3 near page 119 and CW-IMP-1 near page 178 of the PDF; these were located but not read in full.</li>
<li>The source was read selectively. Per-embodiment fields, formal predicates, diagrams, machine-verifiable records, reference software and cross-embodiment combinations in the source were not individually extracted.</li>
<li>The application number and date are applicant-identified; the PDF signature date is 25 October 2026 and is not a certified acknowledgment.</li>
<li>Technical_Disclosure_1.json was not available; Datatracker drafts and vendor documentation were not read; RFC titles are from the editor's knowledge.</li>
<li>No code was executed and no tests or benchmarks are reported.</li>
</ul>
</section>
<section>
<name>Readiness Report and Limitations</name>
<ul>
<li>Section prose, threats, examples, industry rows and platform notes are written by the editor from the work's own task description and general technical knowledge, not extracted from the 3,577 pages.</li>
<li>Overlap with the four earlier drafts must be checked before submission.</li>
<li>Formal IETF IPR disclosure is pending.</li>
</ul>
</section>
<section>
<name>Self-Audit</name>
<ol>
<li>Correct October 25 PDF used (3,577 pages): yes.</li>
<li>All 40 distinct embodiments covered: yes, at the level stated in Appendix A.</li>
<li>Duplicate E31 not counted twice: yes.</li>
<li>Claims 1-60 included: yes (excerpts).</li>
<li>Independent claims 1, 31, 43 and 53 distinguished: yes.</li>
<li>CW-1, CW-2 and CW-3 kept separate: yes.</li>
<li>CW-IMP-1 summarised: yes, from the task description; the source section was located but not read in full.</li>
<li>Interoperability not confused with data exchange: yes.</li>
<li>Source authentication not equated with effect authority: yes.</li>
<li>Permission names not assumed equivalent: yes.</li>
<li>Descriptors carry material properties: yes.</li>
<li>Platform capabilities evaluated realistically: yes.</li>
<li>Observation-only interfaces not called blocking: yes.</li>
<li>Unavailable enforcement leads to non-effective condition: yes.</li>
<li>Task-state movement distinguished from authority transfer: yes.</li>
<li>Source fencing supported by evidence: yes.</li>
<li>Epochs and revocation checked: yes.</li>
<li>Offline limits explicit: yes.</li>
<li>Receipt validation not audit logging: yes.</li>
<li>Earlier success distinguished from unverified acknowledgment: yes.</li>
<li>Human approval does not override other restrictions: yes.</li>
<li>Protected consumption enforced: yes.</li>
<li>Destination reconstruction explicit: yes.</li>
<li>Alternative paths in threat model: yes.</li>
<li>No private Android or iOS APIs invented: yes.</li>
<li>Hardware and software claims match trust assumptions: yes.</li>
<li>Standards credited: yes.</li>
<li>No benchmark invented: yes.</li>
<li>Companies not implied to adopt: yes.</li>
<li>Filing metadata flagged as applicant-identified: yes.</li>
<li>Mothership attribution stated: yes.</li>
<li>Earlier principles not called new in October: yes.</li>
<li>Examiner terms include industry equivalents: yes.</li>
<li>Security and privacy sections present: yes.</li>
<li>IANA section correct: yes.</li>
<li>References resolvable and current: not verified.</li>
<li>IPR without invented licensing: yes.</li>
<li>Useful without patent annexes: yes.</li>
<li>Per-embodiment 20-point analysis: partial (cluster level).</li>
<li>Per-claim 14-field chart: partial (group level).</li>
</ol>
</section>
</section>
</back>
</rfc>
