Verify — "Do the material claims hold?"¶

Mindset: Auditor. RF is a declaration, not a fact. Open files, run necessary checks, compare accepted claims with reality, and state the limits. Test: "Would the evidence establish this claim for this subject, revision and environment without RF?" Map: map.md

Selection Argument¶

Claim IDs	Risk / criticality	Affected behavior / dependencies	Environment	Oracle / authority	Evidence gap / limit	Selected verification and why
{C1…}	{risk}	{scope}	{environment}	{oracle}	{gap or none}	{check / applicable evidence}

Safety/security, human acceptance authority and accepted-result identity are mandatory floors. Expand verification when an observed fact changes a mapped claim, dependency, risk or evidence gap; the existence or count of discrepancies never selects depth by itself.

Verification Log¶

{V1: claim IDs and subject}¶

Accepted claim / authority: {what must hold}
Subject tuple: {accepted subject, revision/Candidate, environment, oracle/authority, dependency state}
Action or evidence: {what was inspected or run}
Observed: {actual result}
Limit: {none or explicit unresolved limit}
Result: HOLDS / FINDING / BLOCKED

Commands Executed¶

#	Command	Claim IDs	Result
1	{build/test/lint/inspection command}	{C1…}	{exit and concise observed result}

Run TS-required checks and checks needed by the selection argument. When reusing evidence, explain why its subject tuple and assumptions still apply. If a required check cannot run, name the exact unresolved material claim and missing environment; this is not PASS. List comments, docstrings and channel words (todo, legacy, old, temp) the Candidate's VALUE diff adds; each must meet conventions.md → Comments.

Claim and Source Checks¶

Spot-check key claims or sources by material reliance, not convenience.
Confirm every citation resolves to a real, relevant artifact.
Verify data claims against a primary source where reachable.

#	Claim / citation	Where	Primary artifact / source	Holds?
C1	{claim}	{file + section}	{source or unresolved}	✅ / ❌ / ⚠️

If the deliverable has no claims, citations or data, state what it does contain and why this is N/A.

Guard and Check Admission¶

#	Kind	Protected behavior / invariant	Failure consequence	Counterfactual detection	Admission
G1	permanent guard / positive control / temporary diagnostic / governance assertion	{subject}	{harm or actual limited use}	{historical red, relevant mutant/fault/fixture, or N/A with reason}	admitted / labelled control / temporary / governance only / rejected

A permanent guard requires the first four fields and relevant demonstrated detection. Historical red-before-green is strong, not universal. A relevant mutant, fault, fixture or equivalent negative control may establish existing behavior. Positive controls, diagnostics, counts, commit/artifact existence and scope arithmetic keep their actual labels and are never product-quality guards by volume.

Candidate Findings¶

A breached criterion, discrepancy, citation or process record becomes verdict-relevant only after the full item contract establishes a material accepted claim or authority consequence.

ID	Class	Subject	Affected claim / authority	Observed fact + oracle	Concrete harm	Material consequence or named absence	Owner	Observable completion	Route / rung	Candidate effect	Status
F1	VALUE / ASSURANCE / TRACE	{subject}	{claim}	{fact + oracle}	{harm}	{material consequence / none}	{owner}	{condition}	{route}	unchanged / moves	open / complete / observation

If empty: No findings. Non-material TRACE remains visible as an observation or finite current-carrier repair; it cannot restart product execution.

Evidence Verification¶

evidence/ holds only EV files; no verdict rests on working material.

#	EV row	Subject tuple	Repeated or opened?	Establishes the claim?	Limit
E1	{EV row ID}	{tuple}	✅ / ❌	✅ / ❌ / ⚠️	{none or limit}

If RF §5 has no evidence items: state N/A — no evidence artifacts to verify and why that can satisfy the mapped claims.

Knowledge Citations Verified¶

Scan PV priorities 0–4 in full and 5–7 by relevance. Verify every HL §7.2 and ONB §7 citation for resolution, item existence, semantic match, currentness and relevance. Check priority 0 purpose and priority 1 methodology separately even when one file contains both.

#	Artifact	Priority + exact citation	Resolves?	Item exists?	Meaning matches?	Relevant?
1	HL §7.2 #{N}	{priority and item}	✅ / ❌	✅ / ❌	✅ / ❌ —	✅ / ❌ —

If HL §7.2 says no applicable items: N/A — no citations to verify.

Accounting Replay¶

Approval / authority	Baseline	Candidate	Literal VALUE membership / actions / classes / reasons	Adds	Deletes	Touched LOC	Binary	Trigger disposition	Exact NUL-safe command	Verdict
{ref and timing}	{full SHA}	{full SHA}	{exact result}	{N}	{N}	{N}	N/A / details	{result}	{command}	VERIFIED / BLOCKED / N/A / INVALID

Candidate is the first tested implementation commit before EV/RF/state. Later VALUE moves it; later TRACE, ASSURANCE or non-value DERIVED does not. Missing/mutable/mismatched/late authority is BLOCKED; metric-only inapplicability is N/A; unresolved attribution is INVALID; DEFERRED is non-terminal.

Selected Knowledge Evidence¶

Inspect contributing-role returns, material sources and completed dispositions against dispatch lineage. Record source/epoch, producer/unit, scope, authority and applicability. Presence, newer time, clean merge or an Applied marker is not independent acceptance.

Checkpoint¶

Self-check: - [ ] Replayed the Map selection and verified all mandatory safety/security, authority and identity floors? - [ ] Established evidence applicability and ran every TS-required or dependency-affected check? - [ ] Recorded explicit limits instead of substituting file, discrepancy, test, commit or artifact counts? - [ ] Classified guards and controls by protected behavior, consequence and counterfactual detection? - [ ] Recorded every candidate finding with the complete item contract and material consequence test? - [ ] Verified RF AC claims, evidence references, citations and immutable accounting against actual artifacts?

Stage complete: YES / NO