@haposoft/cafekit 0.16.0-rc.2 → 0.16.0-rc.3

package/src/claude/scripts/workflow-policy.cjs

@@ -1,15 +1,70 @@
 'use strict';
 
-const REVIEW_VERDICTS = Object.freeze(['PASS', 'FAIL', 'BLOCKED']);
+const crypto = require('crypto');
+const fs = require('fs');
+const path = require('path');
+const PROVENANCE = require('./provenance.cjs');
+
+const CANONICAL_VERDICTS = Object.freeze(['PASS', 'PASS_WITH_WARNINGS', 'FAIL', 'BLOCKED']);
+const LEGACY_DIAGNOSTIC_VERDICTS = Object.freeze(['PARTIAL', 'NO_TESTS']);
+const REVIEW_VERDICTS = CANONICAL_VERDICTS;
 const EXECUTION_TIERS = Object.freeze(['Light', 'Standard', 'Deep']);
 const LANES = Object.freeze(['Direct', 'Standard', 'Critical']);
-const DEEP_TASK_SEQUENCE = Object.freeze(['inspector', 'implementer', 'test-runner', 'code-auditor']);
-const CRITICAL_LANE_SEQUENCE = Object.freeze(['inspector', 'implementer', 'test-runner', 'code-auditor']);
-const LANE_DELEGATION = Object.freeze({
-  Direct: Object.freeze([]),
-  Standard: Object.freeze(['code-auditor']),
-  Critical: CRITICAL_LANE_SEQUENCE,
+const ARTIFACT_PROFILES = Object.freeze(['targeted', 'bounded', 'strict']);
+const PROOF_OBLIGATIONS = Object.freeze([
+  'needsInspection',
+  'needsExecutionProof',
+  'needsIndependentAudit',
+  'needsDurableTaskState',
+  'needsResearchGrounding',
+]);
+const WORKFLOW_POLICY_VERSION = '1';
+const WORKFLOW_POLICY_FIELDS = Object.freeze([
+  'version',
+  'lane',
+  'automatic_lane',
+  'risks',
+  'artifact_profile',
+  'proof_obligations',
+  'actor_needs',
+  'override_receipt',
+]);
+const RECEIPT_BINDING_FIELDS = Object.freeze(['expectedProvenance', 'requireProvenanceBinding']);
+const INDEPENDENT_AUDIT_SCHEMA_VERSION = '1';
+const INDEPENDENT_AUDIT_FIELDS = Object.freeze([
+  'schema_version',
+  'reviewer_session_id',
+  'implementation_session_id',
+  'expected_provenance',
+  'evidence',
+  'verdict',
+]);
+const FLASH_PROMOTION_FIELDS = Object.freeze(['readyForSync', 'flashTransition', 'promotionReceipt']);
+const LEGACY_TIER_LANES = Object.freeze({ Light: 'Direct', Standard: 'Standard', Deep: 'Critical' });
+const ARTIFACT_PROFILE_BY_LANE = Object.freeze({ Direct: 'targeted', Standard: 'bounded', Critical: 'strict' });
+const ACTOR_NEEDS_BY_OBLIGATION = Object.freeze({
+  needsInspection: Object.freeze({ capability: 'inspection', independence: 'same-session' }),
+  needsExecutionProof: Object.freeze({ capability: 'execution-proof', independence: 'same-session' }),
+  needsIndependentAudit: Object.freeze({ capability: 'audit', independence: 'independent' }),
+  needsDurableTaskState: Object.freeze({ capability: 'durable-task-state', independence: 'same-session' }),
+  needsResearchGrounding: Object.freeze({ capability: 'research-grounding', independence: 'same-session' }),
 });
+const ACTOR_CAPABILITIES = new Set(Object.values(ACTOR_NEEDS_BY_OBLIGATION).map((need) => need.capability));
+const ACTOR_INDEPENDENCE = new Set(['same-session', 'independent']);
+const CRITICAL_RISKS = new Set([
+  'reversibility',
+  'destructive',
+  'auth',
+  'payment',
+  'privacy',
+  'data',
+  'schema',
+  'migration',
+  'publicContract',
+  'crossRuntime',
+  'ambiguity',
+  'rollback',
+]);
 const LANE_RISK_KEYS = Object.freeze({
   reversibility: ['reversible', 'reversibility', 'irreversible', 'non-reversible'],
   destructive: ['destructive', 'deletion', 'delete', 'destroy'],
@@ -25,6 +80,342 @@ const LANE_RISK_KEYS = Object.freeze({
   rollback: ['rollback', 'rollback difficulty', 'hard to rollback', 'cannot rollback', 'no rollback'],
 });
 
+function hasOwn(value, key) {
+  return value !== null
+    && (typeof value === 'object' || typeof value === 'function')
+    && Object.prototype.hasOwnProperty.call(value, key);
+}
+
+const PLACEHOLDER_TOKENS = new Set(['TBD','TODO','N/A','NA','NONE','UNKNOWN','PENDING','PLACEHOLDER','REPLACE_ME','-','?']);
+const TAP_METADATA_HEADING_RE = /^\s*#\s*(?:(?:tests?|suites?|pass|fail|cancel(?:l)?ed|skipped|todo)\s+\d+|duration(?:_ms)?\s+\d+(?:\.\d+)?)\s*$/i;
+const ARTIFACT_DECLARATION_KEY = 'artifacts';
+const TASK_ARTIFACT_KEYS = Object.freeze(['artifact', 'artifact_ref', 'artifact_path']);
+const PROVENANCE_VALUE_RE = /^(?:[a-f0-9]{40}|[a-f0-9]{64})$/i;
+const PROSE_RECEIPT_FIELD_RE = /^\s*(?:Notes?|Description|Command(?:\(s\))?)\s*:/i;
+function isPlaceholderToken(value) {
+  if (typeof value !== 'string') return true;
+  const v = value.trim();
+  if (v === '') return true;
+  if (/^\{\{.*\}\}$/.test(v)) return true;
+  if (/^<.*>$/.test(v)) return true;
+  const upper = v.toUpperCase();
+  if (PLACEHOLDER_TOKENS.has(upper)) return true;
+  return false;
+}
+function isTapMetadataHeading(line) {
+  return typeof line === 'string' && TAP_METADATA_HEADING_RE.test(line);
+}
+
+function hasConcreteArtifactDeclaration(value) {
+  if (typeof value === 'string') return !isPlaceholderToken(value);
+  if (Array.isArray(value)) return value.some(hasConcreteArtifactDeclaration);
+  return Boolean(value && typeof value === 'object');
+}
+
+function isSafeArtifactPath(value) {
+  if (typeof value !== 'string' || value !== value.trim() || isPlaceholderToken(value)) return false;
+  if (/^(?:[a-z]:[\\/]|[\\/]|https?:\/\/)/i.test(value)) return false;
+  return !value.split(/[\\/]+/).includes('..');
+}
+
+function isPathInside(root, target) {
+  const relative = path.relative(path.resolve(root), path.resolve(target));
+  return relative !== '' && relative !== '..' && !relative.startsWith(`..${path.sep}`) && !path.isAbsolute(relative);
+}
+
+function stableStat(stat) {
+  return [stat.dev, stat.ino, stat.size, stat.mtimeNs?.toString(), stat.ctimeNs?.toString()].join(':');
+}
+
+function hashValidatedArtifact(root, relativePath, expectedHash) {
+  if (!isSafeArtifactPath(relativePath) || typeof root !== 'string' || root.trim() === '') return false;
+  const rootPath = path.resolve(root);
+  const target = path.resolve(rootPath, relativePath);
+  if (!isPathInside(rootPath, target)) return false;
+
+  const segments = path.relative(rootPath, target).split(path.sep).filter(Boolean);
+  let current = rootPath;
+  try {
+    for (let index = 0; index < segments.length; index += 1) {
+      current = path.join(current, segments[index]);
+      const stat = fs.lstatSync(current);
+      if (stat.isSymbolicLink()) return false;
+      if (index < segments.length - 1 && !stat.isDirectory()) return false;
+    }
+    const rootReal = fs.realpathSync(rootPath);
+    const targetReal = fs.realpathSync(target);
+    if (!isPathInside(rootReal, targetReal)) return false;
+    const before = fs.statSync(target);
+    if (!before.isFile()) return false;
+    const flags = fs.constants.O_RDONLY | (fs.constants.O_NOFOLLOW || 0);
+    const fd = fs.openSync(target, flags);
+    try {
+      const opened = fs.fstatSync(fd);
+      if (!opened.isFile() || stableStat(opened) !== stableStat(before)) return false;
+      const bytes = fs.readFileSync(fd);
+      const after = fs.fstatSync(fd);
+      if (stableStat(after) !== stableStat(opened)) return false;
+      const actual = crypto.createHash('sha256').update(bytes).digest('hex');
+      return actual.toLowerCase() === String(expectedHash).toLowerCase();
+    } finally {
+      fs.closeSync(fd);
+    }
+  } catch (_) {
+    return false;
+  }
+}
+
+function artifactDeclaration(task = {}) {
+  if (Object.prototype.hasOwnProperty.call(task, ARTIFACT_DECLARATION_KEY)) {
+    const artifacts = task[ARTIFACT_DECLARATION_KEY];
+    const unique = Array.isArray(artifacts) ? new Set(artifacts).size === artifacts.length : false;
+    return {
+      declared: true,
+      valid: Array.isArray(artifacts)
+        && artifacts.length > 0
+        && unique
+        && artifacts.every(isSafeArtifactPath),
+      paths: Array.isArray(artifacts) ? [...artifacts] : [],
+    };
+  }
+
+  // Read legacy aliases only for compatibility with pre-P0 task registries.
+  const legacyKey = TASK_ARTIFACT_KEYS.find((key) => hasConcreteArtifactDeclaration(task?.[key]));
+  const legacyValue = legacyKey ? task[legacyKey] : null;
+  const legacyPaths = typeof legacyValue === 'string'
+    ? [legacyValue]
+    : Array.isArray(legacyValue) && legacyValue.every((value) => typeof value === 'string')
+      ? [...legacyValue]
+      : [];
+  return {
+    declared: Boolean(legacyKey),
+    valid: !legacyKey || (legacyPaths.length > 0 && legacyPaths.every(isSafeArtifactPath)),
+    legacy: true,
+    paths: legacyPaths,
+  };
+}
+
+function normalizeProvenanceValue(value) {
+  if (typeof value !== 'string') return null;
+  const normalized = value.trim().replace(/^sha256:/i, '');
+  return normalized || null;
+}
+
+function isConcreteProvenanceValue(value) {
+  return PROVENANCE_VALUE_RE.test(normalizeProvenanceValue(value) || '');
+}
+
+function runtimeContextFromContext(context) {
+  const candidates = [
+    context?.runtime_context,
+    context?.runtimeContext,
+    context?.provenance_context,
+    context?.provenanceContext,
+    context?.receipt_binding?.runtimeContext,
+    context?.receiptBinding?.runtimeContext,
+    context?.provenance_binding?.runtimeContext,
+    context?.provenanceBinding?.runtimeContext,
+  ];
+  const candidate = candidates.find((value) => PROVENANCE.isTrustedRuntimeContext(value));
+  if (!candidate) return null;
+  try { return PROVENANCE.recomputeRuntimeContext(candidate); } catch (_) { return null; }
+}
+
+function firstDefined(source, keys) {
+  if (!source || typeof source !== 'object') return undefined;
+  for (const key of keys) {
+    if (Object.prototype.hasOwnProperty.call(source, key)) return source[key];
+  }
+  return undefined;
+}
+
+function expectedProvenanceFromSource(source) {
+  if (!source || typeof source !== 'object' || Array.isArray(source)) return null;
+  const nested = firstDefined(source, [
+    'expectedProvenance',
+    'expected_provenance',
+  ]);
+  const candidate = nested && typeof nested === 'object' && !Array.isArray(nested)
+    ? nested
+    : source;
+  const base = firstDefined(candidate, [
+    'base', 'Base', 'base_sha', 'baseSha', 'expectedBase', 'expected_base',
+    'expectedBaseSha', 'expected_base_sha',
+  ]);
+  const head = firstDefined(candidate, [
+    'head', 'Head', 'head_sha', 'headSha', 'expectedHead', 'expected_head',
+    'expectedHeadSha', 'expected_head_sha',
+  ]);
+  if (base === undefined && head === undefined) return null;
+  return { base, head };
+}
+
+function createReceiptBinding(runtimeContext) {
+  if (!PROVENANCE.isTrustedRuntimeContext(runtimeContext)) {
+    throw new TypeError('Receipt binding requires a runtime-derived provenance context');
+  }
+  try { return PROVENANCE.createReceiptBinding(runtimeContext); } catch (error) {
+    throw new TypeError(error.message);
+  }
+}
+
+function receiptValidatorOptions(task = {}, {
+  requireProvenanceBinding = false,
+  requireExplicitBinding = false,
+  runtimeContext,
+} = {}) {
+  const declaration = artifactDeclaration(task);
+  let trusted = null;
+  if (runtimeContext !== undefined && PROVENANCE.isTrustedRuntimeContext(runtimeContext)) {
+    try { trusted = PROVENANCE.recomputeRuntimeContext(runtimeContext); } catch (_) { trusted = null; }
+  }
+  return {
+    requireArtifactHash: declaration.declared,
+    artifactDeclarationValid: declaration.valid,
+    artifactPaths: declaration.paths,
+    artifactRoot: trusted?.project_root || null,
+    verifyArtifactBytes: Boolean(trusted?.project_root),
+    expectedProvenance: trusted ? { base: trusted.base, head: trusted.head } : null,
+    requireProvenanceBinding: requireProvenanceBinding || (runtimeContext !== undefined && runtimeContext !== null) || requireExplicitBinding,
+  };
+}
+
+function sha256ValuesFromLine(line) {
+  const values = [];
+  const labels = [...line.matchAll(/\b(?:artifact[_-])?sha-?256\s*:/gi)];
+  for (const label of labels) {
+    const remainder = line.slice(label.index + label[0].length);
+    const token = remainder.match(/^\s*([^\s)\],;`]+)/);
+    values.push(token ? token[1].trim() : '');
+  }
+  return values;
+}
+
+function artifactPathText(line) {
+  return line
+    .replace(/^\s*Artifacts?\s*:\s*/i, '')
+    .replace(/^\s*Artifact\s+produced\s*:?\s*/i, '')
+    .replace(/\b(?:artifact[_-])?sha-?256\s*:\s*[^\s)\],;`]+/gi, '')
+    .replace(/[()[\]{}]/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim()
+    .replace(/(?:\s*[+,;]\s*)+$/, '')
+    .trim();
+}
+
+function parseArtifactDeclarations(body) {
+  const declarations = [];
+  let previous = null;
+  for (const raw of body.split('\n')) {
+    if (PROSE_RECEIPT_FIELD_RE.test(raw)) {
+      previous = null;
+      continue;
+    }
+    const isDeclaration = /^\s*Artifacts?\s*:/i.test(raw)
+      || /^\s*Artifact\s+produced\b/i.test(raw)
+      || /^\s*artifact[_-]?sha-?256\s*:/i.test(raw);
+    if (isDeclaration) {
+      previous = {
+        raw,
+        pathText: artifactPathText(raw),
+        hashes: sha256ValuesFromLine(raw),
+      };
+      declarations.push(previous);
+      continue;
+    }
+    if (previous && /^\s*sha-?256\s*:/i.test(raw)) {
+      previous.hashes.push(...sha256ValuesFromLine(raw));
+      continue;
+    }
+    if (raw.trim() !== '') previous = null;
+  }
+  return declarations;
+}
+
+function artifactSha256Values(body) {
+  return parseArtifactDeclarations(body).flatMap((declaration) => declaration.hashes);
+}
+
+function hasCanonicalArtifactHash(body) {
+  const hashes = artifactSha256Values(body);
+  return hashes.length > 0 && hashes.every((value) => /^[a-f0-9]{64}$/i.test(value));
+}
+
+function hasConcreteArtifactHash(body) {
+  const hashes = artifactSha256Values(body);
+  return hashes.length > 0 && hashes.every((value) => value !== '' && !isPlaceholderToken(value));
+}
+
+function hasExplicitFailure(body) {
+  if (typeof body !== 'string') return false;
+  const lines = body.split('\n');
+  for (const raw of lines) {
+    if (PROSE_RECEIPT_FIELD_RE.test(raw)) continue;
+    const trimmed = raw.trim();
+    if (trimmed === '') continue;
+    // 1. Zero execution (structured, anchored): zero tests, zero passing, or no collection.
+    if (/^\s*(?:#|ℹ)?\s*tests?\s+0\b/i.test(raw)) return true;
+    if (/^\s*(?:#|ℹ)?\s*0\s+tests?\b/i.test(raw)) return true;
+    if (/^\s*(?:#|ℹ)?\s*tests?\s+run\s*:\s*0\b/i.test(raw)) return true;
+    if (/^\s*0\s+(?:passing|passed)\b/i.test(raw)) return true;
+    if (/^\s*Tests?\s*:\s*0\s+(?:passing|passed)\b/i.test(raw)) return true;
+    if (/^\s*Tests?\s+passed\s*:\s*0\b/i.test(raw)) return true;
+    if (/^\s*Passed\s*:\s*0\b/i.test(raw)) return true;
+    if (/^\s*Tests:\s*0\s+total\b/i.test(raw)) return true;
+    if (/^\s*collected\s+0\s+items\b/i.test(raw)) return true;
+    if (/^\s*No tests? found\b/i.test(raw)) return true;
+    // 2. Cancellation (optional marker #/ℹ, cancelled/canceled, both orders, count >0; 0 passes)
+    if (/^\s*(?:#|ℹ)?\s*cancel(?:l)?ed\s*[:\s]*[1-9]\d*\b/i.test(raw)) return true;
+    if (/^\s*(?:#|ℹ)?\s*[1-9]\d*\s+cancel(?:l)?ed\b/i.test(raw)) return true;
+    // 3. Suite / file failures: Test Suites: N failed, Test Files N failed (N>0)
+    if (/^\s*Test Suites?:?\s*[1-9]\d*\s+failed\b/i.test(raw)) return true;
+    if (/^\s*Test Files?:?\s*[1-9]\d*\s+failed\b/i.test(raw)) return true;
+    if (/^\s*Test Suites?:?\s*0\s+failed\b[^\r\n]*\b0\s+passed\b/i.test(raw)) return true;
+    if (/^\s*Test Files?:?\s*0\s+failed\b[^\r\n]*\b0\s+passed\b/i.test(raw)) return true;
+    // 4. Pytest short-summary entries (anchored to a Python test path)
+    if (/^\s*FAILED[ \t]+\S+\.py(?:::\S+)?(?:[ \t]+-[^\r\n]*)?\s*$/i.test(raw)) return true;
+    // 5. Runner FAIL (anchored exact FAIL, horizontal whitespace/colon/end; --- FAIL:)
+    if (/^\s*FAIL(?:[ \t]+|:|$)/.test(raw)) return true;
+    if (/^\s*---\s*FAIL:/.test(raw)) return true;
+    // 6. Error/failure summaries (anchored, N>0; prose Notes/Description/Command skipped)
+    if (/^\s*[1-9]\d*\s+errors?\b/i.test(raw)) return true;
+    if (/^\s*errors?\s*[:\s]*[1-9]\d*\b/i.test(raw)) return true;
+    if (/^\s*failures?\s*[:\s]*[1-9]\d*\b/i.test(raw)) return true;
+    if (/^\s*failure\s+summary\s*[:.]?\s*$/i.test(raw)) return true;
+    if (/^\s*Found\s+[1-9]\d*\s+errors?\b/i.test(raw)) return true;
+    if (/^\s*failureCount\s*:\s*[1-9]\d*\s*[,;]?\s*$/i.test(raw)) return true;
+    if (/^\s*ERROR\s*$/i.test(raw)) return true;
+    if (/^\s*ERROR\s+collecting\b/i.test(raw)) return true;
+    if (/^\s*Tests run\s*:/i.test(raw) && (/\b(?:Failures?|Errors?)\s*[:=]\s*[1-9]/i.test(raw) || /\b0\s+(?:tests?|passing|passed)\b/i.test(raw))) return true;
+    // Maven Surefire's structured failure lines; do not treat arbitrary [ERROR] prose as failure.
+    if (/^\s*\[ERROR\]\s+Tests run:\s*\d+\b[^\r\n]*\b(?:Failures?|Errors?):\s*[1-9]\d*\b/i.test(raw)) return true;
+    if (/^\s*\[ERROR\]\s+Failed to execute goal\b/i.test(raw)) return true;
+    // 7. Gradle (anchored, optional)
+    if (/^\s*\d+\s+tests?\s+completed,\s*[1-9]\d*\s+failed\b/i.test(raw)) return true;
+    // Existing structured checks (preserve, anchored where needed to avoid prose false positive)
+    if (/^\s*Tests?\s+failed\s*$/i.test(raw)) return true;
+    if (/^\s*Tests?\s+failed\s*:\s*[1-9]\d*\b/i.test(raw)) return true;
+    if (/\b(?:Verification|Result|Status|Outcome)\s*:\s*FAIL(?:ED|URE)?\b/i.test(raw)) return true;
+    if (/^\s*FAIL\s*$/i.test(trimmed)) return true;
+    if (/^\s*FAIL\s*:/i.test(trimmed)) return true;
+    if (/^\s*FAILED\s*$/i.test(trimmed)) return true;
+    if (/^\s*FAILED\s*:/i.test(trimmed)) return true;
+    if (/^\s*FAILURE\s*$/i.test(trimmed)) return true;
+    if (/^\s*FAILURE\s*:/i.test(trimmed)) return true;
+    if (/^\s*one\s+failing(?:\s+(?:test|tests|suite|suites))?\s*[,:]?\s*$/i.test(raw)) return true;
+    if (/^\s*(?:#|ℹ)?\s*fail\s+[1-9]/i.test(raw)) return true;
+    if (/^\s*not ok\b/i.test(raw)) return true;
+    if (/^\s*Tests\s*:?\s*[1-9]/i.test(raw) && /\bfailed\b/i.test(raw)) return true;
+    if (/^\s*\d+\s+failed\b/i.test(raw)) {
+      const m = raw.match(/^\s*(\d+)\s+failed\b/i);
+      if (m && parseInt(m[1], 10) !== 0) return true;
+    }
+    if (/exit\s+code\s*[:=]?\s*[1-9]/i.test(raw)) return true;
+  }
+  return false;
+}
+const trustedClassifications = new WeakSet();
+
 const APPROVAL_SCHEMA_VERSION = '2.0';
 const LEGACY_APPROVAL_ERROR = `Legacy approval state detected (field \`approved\`). Migration required: replace \`approved\` with \`agent_validated\` and \`user_approved\` per schema v${APPROVAL_SCHEMA_VERSION} (schema_version: "${APPROVAL_SCHEMA_VERSION}"). See spec-state.json template. Refusing to infer user approval.`;
 
@@ -108,22 +499,37 @@ function overrideLane(input = {}) {
   return input.override ?? input.laneOverride ?? input.requestedLane ?? input.lane ?? null;
 }
 
-function isUserAuthorizedForDowngrade(input = {}) {
-  // Explicit user-originated authorization is required for any downgrade from Critical.
-  // We accept several synonyms for the authorization flag to allow different callers.
-  return input.userAuthorized === true
-    || input.user_authorized === true
-    || input.explicitUserAuthorization === true
-    || input.downgradeAuthorized === true
-    || input.confirmDowngrade === true
-    || input.authorization === 'user'
-    || input.origin === 'user'
-    || input.requestedBy === 'user'
-    || input.user_approved === true;
+function isValidOverrideReceipt(receipt, automaticLane, requestedLane, risks = []) {
+  // P0.4 fail-closed: no trusted runtime issuer exists yet.
+  // Even a correctly shaped object with source:runtime must NOT be considered valid provenance.
+  // Downgrade is blocked until a verifiable, runtime-issued override receipt is implemented
+  // (bound to automaticLane/requestedLane/risks/session and verifiable by that issuer).
+  // Keep this function fail-closed; do not accept plain objects — always return false for P0.
+  return false;
+}
+
+function isUserAuthorizedForDowngrade(input = {}, automaticLane, requestedLane, risks) {
+  // P0.4: downgrade requires structured override receipt with runtime provenance.
+  // Boolean flags (userAuthorized, user_approved, etc.) are never trusted.
+  // Fail closed — no trusted issuer exists yet, so always false.
+  // Explicitly ignore all boolean/string downgrade flags to prevent forged approvals.
+  return false;
+}
+
+function freezeClassification(obj) {
+  Object.freeze(obj);
+  if (Array.isArray(obj.warnings)) Object.freeze(obj.warnings);
+  if (Array.isArray(obj.risks)) Object.freeze(obj.risks);
+  if (obj.signals && typeof obj.signals === 'object') Object.freeze(obj.signals);
+  trustedClassifications.add(obj);
+  return obj;
 }
 
 function classifyLane(input = {}) {
-  if (typeof input === 'string') return { lane: assertLane(input), automaticLane: input, overridden: false, warnings: [], risks: [] };
+  if (typeof input === 'string') {
+    const c = { lane: assertLane(input), automaticLane: input, overridden: false, warnings: [], risks: [], signals: {} };
+    return freezeClassification(c);
+  }
   if (!input || typeof input !== 'object') throw new TypeError('Lane classification input must be an object');
 
   const signals = classifyRiskSignals(input);
@@ -138,7 +544,7 @@ function classifyLane(input = {}) {
       : 'Standard';
   const requested = overrideLane(input);
   if (requested === null || requested === undefined) {
-    return {
+    const c = {
       lane: automaticLane,
       automaticLane,
       overridden: false,
@@ -146,14 +552,15 @@ function classifyLane(input = {}) {
       risks,
       signals,
     };
+    return freezeClassification(c);
   }
   assertLane(requested);
   const isDowngrade = LANES.indexOf(requested) < LANES.indexOf(automaticLane);
-  if (isDowngrade && !isUserAuthorizedForDowngrade(input)) {
+  if (isDowngrade && !isUserAuthorizedForDowngrade(input, automaticLane, requested, risks)) {
     throw new Error(
-      `Downgrade from ${automaticLane} to ${requested} requires explicit user authorization. ` +
+      `Downgrade from ${automaticLane} to ${requested} is blocked: no trusted runtime issuer is available to authorize Critical downgrades. ` +
       `Automatic ${automaticLane} due to: ${risks.join(', ') || 'default risk policy'}. ` +
-      `Provide userAuthorized:true (origin=user) to confirm downgrade and acknowledge reduced review/evidence coverage.`
+      `Keep ${automaticLane} or escalate; downgrade is unavailable until a verifiable runtime override receipt is implemented.`
     );
   }
   const warnings = [`Explicit lane override selected: ${requested}.`];
@@ -161,12 +568,12 @@ function classifyLane(input = {}) {
     warnings.push(`Override changes automatic ${automaticLane} classification to ${requested}.`);
     if (LANES.indexOf(requested) < LANES.indexOf(automaticLane)) {
       warnings.push(`Downgrading risk lane may reduce review and evidence coverage: ${risks.join(', ') || 'default risk policy'}.`);
-      if (isUserAuthorizedForDowngrade(input)) {
+      if (isUserAuthorizedForDowngrade(input, automaticLane, requested, risks)) {
         warnings.push(`Downgrade authorized by user - proceeding with ${requested} despite ${automaticLane} signals.`);
       }
     }
   }
-  return {
+  const c2 = {
     lane: requested,
     automaticLane,
     overridden: true,
@@ -174,29 +581,265 @@ function classifyLane(input = {}) {
     risks,
     signals,
   };
+  return freezeClassification(c2);
 }
 
-function lanePolicy(input = {}) {
-  const classification = typeof input === 'string'
-    ? classifyLane(input)
-    : input && LANES.includes(input.lane) && Object.hasOwn(input, 'automaticLane')
-      ? input
-      : classifyLane(input);
+function uniqueStrings(values, field) {
+  if (!Array.isArray(values)) return { valid: false, error: `${field} must be an array` };
+  if (values.some((value) => typeof value !== 'string' || value.trim() === '')) {
+    return { valid: false, error: `${field} must contain non-empty strings` };
+  }
+  if (new Set(values).size !== values.length) return { valid: false, error: `${field} must not contain duplicates` };
+  return { valid: true };
+}
+
+function obligationsForLane(lane, risks = []) {
+  assertLane(lane);
+  const riskSet = new Set(risks);
+  const obligations = new Set(['needsExecutionProof']);
+  if (lane === 'Standard') obligations.add('needsInspection');
+  if (lane === 'Critical') {
+    obligations.add('needsInspection');
+    obligations.add('needsIndependentAudit');
+    if ([
+      'destructive', 'reversibility', 'rollback', 'data', 'schema', 'migration',
+      'privacy', 'payment', 'crossRuntime',
+    ].some((risk) => riskSet.has(risk))) {
+      obligations.add('needsDurableTaskState');
+    }
+    if ([
+      'auth', 'payment', 'privacy', 'schema', 'migration', 'publicContract',
+      'crossRuntime', 'ambiguity',
+    ].some((risk) => riskSet.has(risk))) {
+      obligations.add('needsResearchGrounding');
+    }
+  }
+  return PROOF_OBLIGATIONS.filter((obligation) => obligations.has(obligation));
+}
+
+function actorNeedsFor(obligations) {
+  return obligations.map((obligation) => ({ ...ACTOR_NEEDS_BY_OBLIGATION[obligation] }));
+}
+
+function minimumLaneForRisks(risks = []) {
+  const listedRisks = Array.isArray(risks) ? risks : [];
+  const classifiedRisks = classifyRiskSignals({ risks: listedRisks });
+  if (Object.entries(classifiedRisks).some(([risk, active]) => active && CRITICAL_RISKS.has(risk))) {
+    return 'Critical';
+  }
+  if (listedRisks.length > 0) return 'Standard';
+  return 'Direct';
+}
+
+function snapshotFromClassification(classification) {
   const lane = assertLane(classification.lane);
-  const delegated = [...LANE_DELEGATION[lane]];
+  const automaticLane = assertLane(classification.automaticLane || lane);
+  if (LANES.indexOf(lane) < LANES.indexOf(automaticLane)) {
+    throw new Error(`Workflow policy cannot downgrade ${automaticLane} to ${lane} without a trusted override receipt`);
+  }
+  const risks = Array.isArray(classification.risks) ? [...new Set(classification.risks)] : [];
+  const minimumLane = minimumLaneForRisks(risks);
+  if (LANES.indexOf(lane) < LANES.indexOf(minimumLane)) {
+    throw new Error(`Workflow policy lane ${lane} is below ${minimumLane} for risks: ${risks.join(', ')}`);
+  }
   return {
+    version: WORKFLOW_POLICY_VERSION,
     lane,
-    executionTier: lane === 'Critical' ? 'Deep' : lane === 'Direct' ? 'Light' : 'Standard',
-    automaticLane: classification.automaticLane,
-    overridden: classification.overridden,
-    warnings: [...classification.warnings],
-    risks: [...classification.risks],
-    delegated,
+    automatic_lane: automaticLane,
+    risks,
+    artifact_profile: ARTIFACT_PROFILE_BY_LANE[lane],
+    proof_obligations: obligationsForLane(lane, risks),
+    actor_needs: actorNeedsFor(obligationsForLane(lane, risks)),
+    override_receipt: null,
+  };
+}
+
+function workflowPolicySnapshot(input = {}) {
+  if (input && typeof input === 'object' && hasOwn(input, 'workflow_policy')) {
+    return readWorkflowPolicySnapshot(input);
+  }
+  if (input && typeof input === 'object' && input.design_context?.execution_tier) {
+    return readWorkflowPolicySnapshot(input);
+  }
+  if (input && typeof input === 'object' && isWorkflowPolicySnapshot(input)) {
+    assertWorkflowPolicySnapshot(input);
+    return JSON.parse(JSON.stringify(input));
+  }
+  const classification = trustedClassifications.has(input)
+    ? input
+    : classifyLane(input);
+  return snapshotFromClassification(classification);
+}
+
+function isWorkflowPolicySnapshot(value) {
+  return Boolean(value && typeof value === 'object' && Object.prototype.hasOwnProperty.call(value, 'version')
+    && Object.prototype.hasOwnProperty.call(value, 'automatic_lane')
+    && Object.prototype.hasOwnProperty.call(value, 'artifact_profile'));
+}
+
+function validateWorkflowPolicySnapshot(snapshot) {
+  const errors = [];
+  if (!snapshot || typeof snapshot !== 'object' || Array.isArray(snapshot)) {
+    return { valid: false, errors: ['workflow_policy must be an object'] };
+  }
+  const keys = Object.keys(snapshot).sort();
+  const expectedKeys = [...WORKFLOW_POLICY_FIELDS].sort();
+  if (JSON.stringify(keys) !== JSON.stringify(expectedKeys)) {
+    errors.push(`workflow_policy fields must be exactly ${WORKFLOW_POLICY_FIELDS.join(', ')}`);
+  }
+  if (snapshot.version !== WORKFLOW_POLICY_VERSION) errors.push('workflow_policy.version must be "1"');
+  if (!LANES.includes(snapshot.lane)) errors.push(`workflow_policy.lane must be one of ${LANES.join(', ')}`);
+  if (!LANES.includes(snapshot.automatic_lane)) errors.push(`workflow_policy.automatic_lane must be one of ${LANES.join(', ')}`);
+  if (LANES.includes(snapshot.lane) && LANES.includes(snapshot.automatic_lane)
+    && LANES.indexOf(snapshot.lane) < LANES.indexOf(snapshot.automatic_lane)) {
+    errors.push('workflow_policy.lane cannot be lower than automatic_lane without a trusted override receipt');
+  }
+  if (LANES.includes(snapshot.lane) && snapshot.artifact_profile !== ARTIFACT_PROFILE_BY_LANE[snapshot.lane]) {
+    errors.push(`workflow_policy.artifact_profile must be ${ARTIFACT_PROFILE_BY_LANE[snapshot.lane]} for ${snapshot.lane}`);
+  }
+  const risks = uniqueStrings(snapshot.risks, 'workflow_policy.risks');
+  if (!risks.valid) errors.push(risks.error);
+  if (risks.valid && LANES.includes(snapshot.lane)) {
+    const minimumLane = minimumLaneForRisks(snapshot.risks);
+    if (LANES.indexOf(snapshot.lane) < LANES.indexOf(minimumLane)) {
+      errors.push(`workflow_policy.lane must be at least ${minimumLane} for its risks`);
+    }
+  }
+  const obligations = uniqueStrings(snapshot.proof_obligations, 'workflow_policy.proof_obligations');
+  if (!obligations.valid) {
+    errors.push(obligations.error);
+  } else {
+    for (const obligation of snapshot.proof_obligations) {
+      if (!PROOF_OBLIGATIONS.includes(obligation)) errors.push(`workflow_policy.proof_obligations has unknown obligation ${obligation}`);
+    }
+    if (!snapshot.proof_obligations.includes('needsExecutionProof')) {
+      errors.push('workflow_policy.proof_obligations must include needsExecutionProof');
+    }
+    if (LANES.includes(snapshot.lane) && risks.valid) {
+      const expected = obligationsForLane(snapshot.lane, snapshot.risks);
+      if (JSON.stringify(snapshot.proof_obligations) !== JSON.stringify(expected)) {
+        errors.push('workflow_policy.proof_obligations do not match lane and risks');
+      }
+    }
+  }
+  if (!Array.isArray(snapshot.actor_needs)) {
+    errors.push('workflow_policy.actor_needs must be an array');
+  } else {
+    for (const need of snapshot.actor_needs) {
+      if (!need || typeof need !== 'object' || Array.isArray(need)
+        || Object.keys(need).sort().join(',') !== 'capability,independence'
+        || !ACTOR_CAPABILITIES.has(need.capability)
+        || !ACTOR_INDEPENDENCE.has(need.independence)) {
+        errors.push('workflow_policy.actor_needs must describe capability and independence, never an agent name');
+        break;
+      }
+    }
+    if (LANES.includes(snapshot.lane) && obligations.valid
+      && JSON.stringify(snapshot.actor_needs) !== JSON.stringify(actorNeedsFor(snapshot.proof_obligations))) {
+      errors.push('workflow_policy.actor_needs must correspond to proof_obligations');
+    }
+  }
+  if (snapshot.override_receipt !== null) {
+    errors.push('workflow_policy.override_receipt must be null until a trusted downgrade issuer exists');
+  }
+  return { valid: errors.length === 0, errors };
+}
+
+function assertWorkflowPolicySnapshot(snapshot) {
+  const result = validateWorkflowPolicySnapshot(snapshot);
+  if (!result.valid) throw new TypeError(`Invalid workflow policy snapshot: ${result.errors.join('; ')}`);
+  return snapshot;
+}
+
+function readWorkflowPolicySnapshot(spec = {}) {
+  if (!spec || typeof spec !== 'object') throw new TypeError('Spec state must be an object');
+  if (hasOwn(spec, 'workflow_policy')) {
+    assertWorkflowPolicySnapshot(spec.workflow_policy);
+    return JSON.parse(JSON.stringify(spec.workflow_policy));
+  }
+  // Legacy state is read-compatible only. The fallback is deliberately not
+  // written back and cannot authorize a downgrade or override new input.
+  const rawLegacyTier = spec.design_context && spec.design_context.execution_tier;
+  const legacyTier = typeof rawLegacyTier === 'string'
+    ? Object.keys(LEGACY_TIER_LANES).find((tier) => tier.toLowerCase() === rawLegacyTier.toLowerCase())
+    : null;
+  const legacyLane = legacyTier ? LEGACY_TIER_LANES[legacyTier] : null;
+  if (legacyLane) return snapshotFromClassification({ lane: legacyLane, automaticLane: legacyLane, risks: [] });
+  throw new TypeError('workflow_policy snapshot is missing at the spec boundary');
+}
+
+function persistWorkflowPolicySnapshot(spec, input = {}) {
+  if (!spec || typeof spec !== 'object' || Array.isArray(spec)) throw new TypeError('Spec state must be an object');
+  const next = { ...spec };
+  delete next.override_receipt;
+  if (hasOwn(spec, 'workflow_policy')) {
+    assertWorkflowPolicySnapshot(spec.workflow_policy);
+    return { ...next, workflow_policy: JSON.parse(JSON.stringify(spec.workflow_policy)) };
+  }
+  return { ...next, workflow_policy: workflowPolicySnapshot(input) };
+}
+
+const persistWorkflowPolicy = persistWorkflowPolicySnapshot;
+const createWorkflowPolicySnapshot = workflowPolicySnapshot;
+
+function candidateLaneForRisks(input, risks) {
+  const explicitLevel = String(input?.riskLevel || input?.risk_level || input?.level || '').toLowerCase();
+  const explicitLane = ['critical', 'high', 'deep'].includes(explicitLevel)
+    ? 'Critical'
+    : ['standard', 'medium'].includes(explicitLevel)
+      ? 'Standard'
+      : null;
+  const riskLane = minimumLaneForRisks(risks);
+  if (explicitLane === null && riskLane === 'Direct') return null;
+  if (explicitLane === null) return riskLane;
+  return LANES[Math.max(LANES.indexOf(explicitLane), LANES.indexOf(riskLane))];
+}
+
+function escalateWorkflowPolicy(policyOrSpec, riskInput = {}) {
+  const current = policyOrSpec && typeof policyOrSpec === 'object' && hasOwn(policyOrSpec, 'workflow_policy')
+    ? readWorkflowPolicySnapshot(policyOrSpec)
+    : workflowPolicySnapshot(policyOrSpec);
+  const discovered = classifyRiskSignals(riskInput);
+  const explicitRisks = Array.isArray(riskInput?.risks)
+    ? riskInput.risks.filter((risk) => typeof risk === 'string' && risk.trim() !== '')
+    : [];
+  const newRisks = [...new Set([
+    ...Object.entries(discovered).filter(([, active]) => active).map(([name]) => name),
+    ...explicitRisks,
+  ])];
+  const risks = [...new Set([...current.risks, ...newRisks])];
+  if (newRisks.every((risk) => current.risks.includes(risk))) return current;
+  const candidate = candidateLaneForRisks(riskInput, newRisks);
+  if (!candidate) return current;
+  const lane = LANES[Math.max(LANES.indexOf(current.lane), LANES.indexOf(candidate))];
+  return snapshotFromClassification({
+    lane,
+    automaticLane: current.automatic_lane,
+    risks,
+  });
+}
+
+function policyForSpec(spec, riskInput = null) {
+  const snapshot = riskInput ? escalateWorkflowPolicy(spec, riskInput) : readWorkflowPolicySnapshot(spec);
+  return lanePolicy(snapshot);
+}
+
+function lanePolicy(input = {}) {
+  const snapshot = input && typeof input === 'object' && hasOwn(input, 'workflow_policy')
+    ? readWorkflowPolicySnapshot(input)
+    : input && typeof input === 'object' && isWorkflowPolicySnapshot(input)
+      ? (assertWorkflowPolicySnapshot(input), JSON.parse(JSON.stringify(input)))
+      : workflowPolicySnapshot(input);
+  const lane = assertLane(snapshot.lane);
+  return {
+    ...snapshot,
+    proof_obligations: [...snapshot.proof_obligations],
+    actor_needs: snapshot.actor_needs.map((need) => ({ ...need })),
     requiresSpec: lane !== 'Direct',
-    requiresState: lane !== 'Direct',
+    requiresState: lane === 'Critical',
     qualityGate: lane === 'Direct' ? 'main-session' : lane === 'Standard' ? 'combined-feature-review' : 'strict-evidence',
     shipPoint: lane === 'Direct' ? 'task' : 'feature',
-    evidence: lane === 'Critical' ? 'strict' : lane === 'Standard' ? 'bounded' : 'targeted',
+    evidence: snapshot.artifact_profile,
   };
 }
 
@@ -205,85 +848,106 @@ function isLegacyApprovalObject(obj) {
 }
 
 function validateApprovalSchema(spec) {
-  if (!spec || typeof spec !== 'object') return { valid: false, legacy: false, error: 'spec must be an object' };
+  if (!spec || typeof spec !== 'object' || Array.isArray(spec)) return { valid: false, absent: false, legacy: false, error: 'spec must be a non-array object' };
+  if (!Object.prototype.hasOwnProperty.call(spec, 'approvals')) {
+    return { valid: true, absent: true, legacy: false };
+  }
   const approvals = spec.approvals;
-  if (!approvals || typeof approvals !== 'object') return { valid: true, legacy: false };
+  if (approvals === null || typeof approvals !== 'object' || Array.isArray(approvals)) {
+    return { valid: false, absent: false, legacy: false, error: 'approvals must be a non-empty object' };
+  }
+  if (Object.keys(approvals).length === 0) {
+    return { valid: false, absent: false, legacy: false, error: 'approvals must not be empty' };
+  }
   // Check for legacy field `approved` in any stage
   for (const [stage, val] of Object.entries(approvals)) {
     if (isLegacyApprovalObject(val)) {
       return {
         valid: false,
+        absent: false,
         legacy: true,
         error: `Legacy approval state at approvals.${stage}.approved. ${LEGACY_APPROVAL_ERROR}`,
       };
     }
-    if (val && typeof val === 'object') {
-      // also check if missing new fields but has generated/approved only
-      if ('generated' in val && !('agent_validated' in val) && !('user_approved' in val) && !('approved' in val)) {
-        // This is ambiguous: generated without new fields - treat as not ready but not legacy
-      }
+    if (!val || typeof val !== 'object' || Array.isArray(val)) {
+      return { valid: false, absent: false, legacy: false, error: `approvals.${stage} must be an approval object` };
+    }
+    const keys = Object.keys(val).sort();
+    if (keys.join(',') !== 'agent_validated,generated,user_approved') {
+      return { valid: false, absent: false, legacy: false, error: `approvals.${stage} must contain exactly generated, agent_validated, and user_approved` };
+    }
+    if (keys.some((key) => typeof val[key] !== 'boolean')) {
+      return { valid: false, absent: false, legacy: false, error: `approvals.${stage} fields must be boolean` };
     }
   }
   // Check schema version
-  const version = spec.schema_version || spec.approval_schema_version;
+  const version = Object.prototype.hasOwnProperty.call(spec, 'schema_version')
+    ? spec.schema_version
+    : spec.approval_schema_version;
+  if (version !== undefined && typeof version !== 'string') {
+    return { valid: false, absent: false, legacy: false, error: 'schema_version must be a string' };
+  }
   if (version && version !== APPROVAL_SCHEMA_VERSION) {
-    return { valid: false, legacy: false, error: `Unsupported schema_version ${version}, expected ${APPROVAL_SCHEMA_VERSION}` };
+    return { valid: false, absent: false, legacy: false, error: `Unsupported schema_version ${version}, expected ${APPROVAL_SCHEMA_VERSION}` };
   }
   if (!version && Object.keys(approvals).length > 0) {
     // No version but uses new schema - allow for backward compat but warn that version should be present
     // Fail closed only if legacy detected above; otherwise allow.
-    return { valid: true, legacy: false, warning: `Missing schema_version, expected ${APPROVAL_SCHEMA_VERSION}` };
+    return { valid: true, absent: false, legacy: false, warning: `Missing schema_version, expected ${APPROVAL_SCHEMA_VERSION}` };
   }
-  return { valid: true, legacy: false };
+  return { valid: true, absent: false, legacy: false };
 }
 
 function approvalState(state = {}) {
-  // Detect legacy at top level or inside approvals map
-  const approvalsMap = state && state.approvals && typeof state.approvals === 'object' ? state.approvals : null;
-  if (approvalsMap) {
-    for (const val of Object.values(approvalsMap)) {
-      if (isLegacyApprovalObject(val)) {
-        throw new Error(LEGACY_APPROVAL_ERROR);
-      }
-    }
+  if (!state || typeof state !== 'object' || Array.isArray(state)) {
+    throw new TypeError('approval state must be an object');
+  }
+  if (Object.prototype.hasOwnProperty.call(state, 'approvals')) {
+    const validation = validateApprovalSchema(state);
+    if (!validation.valid) throw new Error(validation.error);
+    return {
+      generated: false,
+      agent_validated: false,
+      user_approved: false,
+      ready: false,
+      present: true,
+      schema_version: APPROVAL_SCHEMA_VERSION,
+      legacy: false,
+    };
+  }
+  if (isLegacyApprovalObject(state)) throw new Error(LEGACY_APPROVAL_ERROR);
+
+  const version = Object.prototype.hasOwnProperty.call(state, 'schema_version')
+    ? state.schema_version
+    : state.approval_schema_version;
+  if (version !== undefined && (typeof version !== 'string' || version !== APPROVAL_SCHEMA_VERSION)) {
+    throw new Error(`Unsupported schema_version ${String(version)}, expected ${APPROVAL_SCHEMA_VERSION}`);
   }
-  const source = approvalsMap ? null : state;
-  // If source is approvals map itself, we should not try to interpret it as single stage
-  // But approvalState is intended for a single stage object; if passed whole approvals map, treat as error?
-  // Keep original behavior: if state has approvals, use that, otherwise state itself.
-  // For whole-spec case with legacy inside map, we already threw above.
-  // For single-stage object:
-  const singleSource = state && state.approvals && typeof state.approvals === 'object' && !approvalsMap ? state.approvals : state;
-  // Actually simplify: if state has approvals, we already handled legacy; for single stage, source is state itself.
-  // If state was {generated:true, agent_validated:true}, then approvalsMap is null, source = state
-  const target = approvalsMap ? null : state;
-  if (target && isLegacyApprovalObject(target)) {
-    throw new Error(LEGACY_APPROVAL_ERROR);
-  }
-  // For single stage case, use target
-  // For map case, we shouldn't be called with map; but if called with spec object containing approvals map, we return not-ready?
-  if (approvalsMap) {
-    // This is a spec-level object, not a stage. Return aggregate? For compatibility, treat as not ready.
-    // Check if any stage has valid approval? But spec-level ready requires all stages.
-    // We'll just return not ready with legacy check already done.
-    return { generated: false, agent_validated: false, user_approved: false, ready: false, schema_version: APPROVAL_SCHEMA_VERSION, legacy: false };
-  }
-  const src = target || state;
-  // Final legacy check on src
-  if (isLegacyApprovalObject(src)) {
-    throw new Error(LEGACY_APPROVAL_ERROR);
+
+  const known = ['generated', 'agent_validated', 'user_approved'];
+  const present = known.some((key) => Object.prototype.hasOwnProperty.call(state, key));
+  const unknown = Object.keys(state).filter((key) => !known.includes(key) && key !== 'schema_version' && key !== 'approval_schema_version');
+  if (unknown.length > 0) throw new TypeError(`malformed approval state fields: ${unknown.join(', ')}`);
+  for (const key of known) {
+    if (Object.prototype.hasOwnProperty.call(state, key) && typeof state[key] !== 'boolean') {
+      throw new TypeError(`approval state field ${key} must be boolean`);
+    }
   }
   const result = {
-    generated: src?.generated === true,
-    agent_validated: src?.agent_validated === true,
-    user_approved: src?.user_approved === true,
+    generated: state.generated === true,
+    agent_validated: state.agent_validated === true,
+    user_approved: state.user_approved === true,
+  };
+  return {
+    ...result,
+    ready: result.generated && result.agent_validated && result.user_approved,
+    present,
+    schema_version: APPROVAL_SCHEMA_VERSION,
   };
-  return { ...result, ready: result.generated && result.agent_validated && result.user_approved, schema_version: APPROVAL_SCHEMA_VERSION };
 }
-
 function assertVerdict(verdict) {
-  if (!REVIEW_VERDICTS.includes(verdict)) {
-    throw new TypeError(`Unsupported review verdict: ${String(verdict)}`);
+  if (!CANONICAL_VERDICTS.includes(verdict)) {
+    throw new TypeError(`Unsupported canonical verdict: ${String(verdict)}`);
   }
   return verdict;
 }
@@ -310,10 +974,9 @@ function delegationPlan({ tier, lane = null, mode = 'full-spec', taskCount = 1 }
     const policy = lanePolicy(lane);
     return {
       lane: policy.lane,
-      executionTier: policy.executionTier,
-      tier: tier || null,
       mode,
-      delegated: policy.delegated,
+      proof_obligations: [...policy.proof_obligations],
+      actor_needs: policy.actor_needs.map((need) => ({ ...need })),
       qualityGate: policy.qualityGate,
       shipPoint: policy.shipPoint,
     };
@@ -325,65 +988,476 @@ function delegationPlan({ tier, lane = null, mode = 'full-spec', taskCount = 1 }
     throw new RangeError('taskCount must be a positive integer');
   }
 
-  if (tier === 'Light') {
-    return { tier, mode, delegated: [], qualityGate: 'main-session', shipPoint: 'task' };
+  // Legacy tier input is read-only compatibility. It is translated once to
+  // the canonical lane contract and never emits a new tier authority or an
+  // agent sequence.
+  const legacyLane = LEGACY_TIER_LANES[tier];
+  const policy = lanePolicy(legacyLane);
+  return {
+    tier,
+    legacy: true,
+    lane: policy.lane,
+    mode,
+    proof_obligations: [...policy.proof_obligations],
+    actor_needs: policy.actor_needs.map((need) => ({ ...need })),
+    qualityGate: policy.qualityGate,
+    shipPoint: policy.shipPoint,
+    taskCount,
+  };
+}
+
+function normalizeVerdict(verdict) {
+  if (CANONICAL_VERDICTS.includes(verdict)) return { verdict, diagnostic: null };
+  if (LEGACY_DIAGNOSTIC_VERDICTS.includes(verdict)) return { verdict: 'BLOCKED', diagnostic: verdict };
+  throw new TypeError(`Unsupported canonical verdict: ${String(verdict)}`);
+}
+
+function unfinishedVerdictDecision(verdict, context = {}, diagnostic = null) {
+  const action = diagnostic === 'PARTIAL'
+    ? 'continue-verification'
+    : diagnostic === 'NO_TESTS'
+      ? 'configure-or-run-tests'
+      : verdict === 'FAIL'
+        ? 'fix-and-rerun'
+        : 'stop';
+  return {
+    verdict,
+    ...(diagnostic ? { diagnostic } : {}),
+    action,
+    completion: 'unfinished',
+    terminal: verdict !== 'FAIL',
+    unfinished: true,
+    retry: verdict === 'FAIL',
+    blocker: context.blocker || (
+      diagnostic === 'PARTIAL'
+        ? 'verification is partial; required proof remains unfinished'
+        : diagnostic === 'NO_TESTS'
+          ? 'no tests were executed'
+          : verdict === 'FAIL'
+            ? 'verification returned FAIL'
+            : 'verification is blocked; resolve the blocker before retrying'
+    ),
+  };
+}
+
+function adaptVerdict(verdict, context = {}) {
+  const normalized = normalizeVerdict(verdict);
+  if (normalized.verdict === 'PASS' || normalized.verdict === 'PASS_WITH_WARNINGS') {
+    return {
+      verdict: normalized.verdict,
+      ...(normalized.diagnostic ? { diagnostic: normalized.diagnostic } : {}),
+      action: normalized.verdict === 'PASS_WITH_WARNINGS' ? 'review-passed-with-warnings' : 'review-passed',
+      completion: 'unfinished',
+      terminal: true,
+      unfinished: true,
+      retry: false,
+      blocker: context.blocker || 'completion decision requires canonical execution proof and workflow obligations',
+    };
+  }
+  return unfinishedVerdictDecision(normalized.verdict, context, normalized.diagnostic);
+}
+
+function workflowPolicyObligations(context) {
+  if (!context || typeof context !== 'object' || Array.isArray(context)) return null;
+  const policy = hasOwn(context, 'workflow_policy')
+    ? context.workflow_policy
+    : hasOwn(context, 'workflowPolicy')
+      ? context.workflowPolicy
+      : hasOwn(context, 'policy')
+        ? context.policy
+        : context.task_context && typeof context.task_context === 'object'
+          ? (hasOwn(context.task_context, 'workflow_policy')
+            ? context.task_context.workflow_policy
+            : hasOwn(context.task_context, 'workflowPolicy')
+              ? context.task_context.workflowPolicy
+              : null)
+          : null;
+  if (!validateWorkflowPolicySnapshot(policy).valid) return null;
+  return [...policy.proof_obligations];
+}
+
+function executionReceiptFromContext(context) {
+  const candidates = [
+    context?.execution_receipt,
+    context?.executionReceipt,
+    context?.canonical_execution_receipt,
+    context?.canonicalExecutionReceipt,
+    context?.execution?.receipt,
+    context?.execution?.body,
+    context?.receipt,
+  ];
+  return candidates.find((value) => typeof value === 'string' && value.trim() !== '') || null;
+}
+
+function proofContainers(context) {
+  return [
+    context?.proofs,
+    context?.workflow_proofs,
+    context?.workflowProofs,
+    context?.evidence,
+    context?.obligation_evidence,
+    context?.obligationEvidence,
+  ].filter((value) => value && typeof value === 'object' && !Array.isArray(value));
+}
+
+function concreteEvidence(value) {
+  if (typeof value === 'string') return value.trim() !== '' && !isPlaceholderToken(value);
+  if (Array.isArray(value)) return value.some(concreteEvidence);
+  if (!value || typeof value !== 'object') return false;
+  return ['evidence', 'proof', 'receipt', 'report', 'summary']
+    .some((key) => concreteEvidence(value[key]));
+}
+
+function evidenceForObligation(context, obligation) {
+  for (const container of proofContainers(context)) {
+    if (Object.prototype.hasOwnProperty.call(container, obligation) && concreteEvidence(container[obligation])) {
+      return container[obligation];
+    }
+  }
+  const aliases = {
+    needsInspection: ['inspection', 'inspectionEvidence', 'inspection_evidence'],
+    needsDurableTaskState: ['durableTaskState', 'durable_task_state', 'taskState', 'task_state'],
+    needsResearchGrounding: ['research', 'researchGrounding', 'research_grounding'],
+  };
+  for (const key of aliases[obligation] || []) {
+    if (concreteEvidence(context?.[key])) return context[key];
   }
-  if (tier === 'Standard') {
+  return null;
+}
+
+function independentAuditEvidence(context) {
+  const runtimeContext = runtimeContextFromContext(context);
+  const expectedProvenance = runtimeContext
+    ? { base: runtimeContext.base, head: runtimeContext.head }
+    : null;
+  const candidates = [
+    ...proofContainers(context).map((container) => container.needsIndependentAudit),
+    context?.independentAudit,
+    context?.independent_audit,
+    context?.audit,
+  ];
+  return candidates.find((value) => {
+    if (Array.isArray(value)) return false;
+    return validateIndependentAuditEvidence(value, expectedProvenance, runtimeContext).valid;
+  }) || null;
+}
+
+function validateIndependentAuditEvidence(value, expectedProvenance, runtimeContext) {
+  const errors = [];
+  if (!value || typeof value !== 'object' || Array.isArray(value)) {
+    return { valid: false, errors: ['audit must be an object'] };
+  }
+  if (JSON.stringify(Object.keys(value).sort()) !== JSON.stringify([...INDEPENDENT_AUDIT_FIELDS].sort())) {
+    errors.push(`audit fields must be exactly ${INDEPENDENT_AUDIT_FIELDS.join(', ')}`);
+  }
+  if (value.schema_version !== INDEPENDENT_AUDIT_SCHEMA_VERSION) {
+    errors.push(`audit.schema_version must be "${INDEPENDENT_AUDIT_SCHEMA_VERSION}"`);
+  }
+  for (const field of ['reviewer_session_id', 'implementation_session_id']) {
+    if (typeof value[field] !== 'string' || isPlaceholderToken(value[field])) {
+      errors.push(`audit.${field} must be concrete`);
+    }
+  }
+  if (typeof value.reviewer_session_id === 'string'
+    && typeof value.implementation_session_id === 'string'
+    && value.reviewer_session_id.trim() === value.implementation_session_id.trim()) {
+    errors.push('audit reviewer and implementation sessions must be distinct');
+  }
+  if (typeof value.evidence !== 'string' || isPlaceholderToken(value.evidence)) {
+    errors.push('audit.evidence must be concrete');
+  }
+  if (value.verdict !== 'PASS') errors.push('audit.verdict must be literal PASS');
+
+  if (!PROVENANCE.isTrustedRuntimeContext(runtimeContext)) {
+    errors.push('audit runtime provenance/session context is unavailable');
+  } else if (typeof value.implementation_session_id === 'string'
+    && value.implementation_session_id.trim() !== runtimeContext.runtime_session) {
+    errors.push('audit implementation session does not match the runtime session');
+  }
+
+  if (!value.expected_provenance
+    || typeof value.expected_provenance !== 'object'
+    || Array.isArray(value.expected_provenance)
+    || JSON.stringify(Object.keys(value.expected_provenance).sort()) !== JSON.stringify(['base', 'head'])) {
+    errors.push('audit.expected_provenance must contain exactly base and head');
+  }
+  const actualBinding = expectedProvenanceFromSource(value.expected_provenance);
+  const runtimeBinding = PROVENANCE.isTrustedRuntimeContext(runtimeContext)
+    ? { base: runtimeContext.base, head: runtimeContext.head }
+    : null;
+  if (!actualBinding
+    || !isConcreteProvenanceValue(actualBinding.base)
+    || !isConcreteProvenanceValue(actualBinding.head)
+    || !expectedProvenance
+    || !isConcreteProvenanceValue(expectedProvenance.base)
+    || !isConcreteProvenanceValue(expectedProvenance.head)
+    || normalizeProvenanceValue(actualBinding.base) !== normalizeProvenanceValue(expectedProvenance.base)
+    || normalizeProvenanceValue(actualBinding.head) !== normalizeProvenanceValue(expectedProvenance.head)) {
+    errors.push('audit expected Base and Head binding does not match the runtime binding');
+  }
+  if (runtimeBinding
+    && (normalizeProvenanceValue(actualBinding?.base) !== runtimeBinding.base
+      || normalizeProvenanceValue(actualBinding?.head) !== runtimeBinding.head)) {
+    errors.push('audit expected Base and Head do not match the freshly derived runtime binding');
+  }
+  return { valid: errors.length === 0, errors };
+}
+
+function completionReceiptOptions(context) {
+  const task = context?.task_context || context?.taskContext || context?.task || context;
+  const runtimeContext = runtimeContextFromContext(context);
+  const derived = receiptValidatorOptions(task && typeof task === 'object' ? task : {}, {
+    requireProvenanceBinding: true,
+    runtimeContext,
+  });
+  const supplied = context?.receipt_options || context?.receiptOptions;
+  if (!supplied || typeof supplied !== 'object' || Array.isArray(supplied)) return derived;
+  return {
+    ...derived,
+    ...supplied,
+    artifactDeclarationValid: derived.artifactDeclarationValid === false
+      ? false
+      : supplied.artifactDeclarationValid !== false,
+    requireArtifactHash: derived.requireArtifactHash === true || supplied.requireArtifactHash === true,
+    artifactPaths: derived.artifactPaths.length > 0 ? derived.artifactPaths : (Array.isArray(supplied.artifactPaths) ? supplied.artifactPaths : []),
+    artifactRoot: derived.artifactRoot,
+    verifyArtifactBytes: derived.verifyArtifactBytes === true || supplied.verifyArtifactBytes === true,
+    expectedProvenance: derived.expectedProvenance,
+    requireProvenanceBinding: true,
+  };
+}
+
+function runtimeBoundProofsFromSpec(spec, runtimeContext) {
+  // Spec files are worker-writable: plain proof strings are never execution
+  // evidence. Only canonical receipts bound to this fresh runtime may enter
+  // the completion decision; independent audit remains a strict object schema.
+  const proofSources = [
+    'completion_proofs',
+    'completionProofs',
+    'workflow_proofs',
+    'workflowProofs',
+    'obligation_evidence',
+    'obligationEvidence',
+    'proofs',
+  ];
+  const sourceKey = proofSources.find((key) => spec && spec[key] && typeof spec[key] === 'object' && !Array.isArray(spec[key]));
+  if (!sourceKey) return null;
+  const source = spec[sourceKey];
+  const trusted = {};
+  const receiptOptions = receiptValidatorOptions({}, {
+    runtimeContext,
+    requireProvenanceBinding: true,
+  });
+  for (const [obligation, value] of Object.entries(source)) {
+    if (obligation === 'needsIndependentAudit') {
+      trusted[obligation] = value;
+      continue;
+    }
+    const candidates = value && typeof value === 'object' && !Array.isArray(value)
+      ? [value.receipt, value.execution_receipt, value.executionReceipt]
+      : [];
+    const receipt = candidates.find((candidate) => (
+      typeof candidate === 'string'
+      && validateCanonicalReceipt(candidate, receiptOptions).length === 0
+    ));
+    if (receipt) trusted[obligation] = { receipt };
+  }
+  return trusted;
+}
+
+function completionDecisionForSpec(spec, {
+  runtimeContext,
+  executionReceipt,
+  taskContext,
+} = {}) {
+  const policy = spec && typeof spec === 'object' ? spec.workflow_policy : null;
+  const context = {
+    workflow_policy: policy,
+    runtime_context: runtimeContext,
+    execution_receipt: executionReceipt,
+    task_context: taskContext,
+  };
+  const trustedProofs = runtimeBoundProofsFromSpec(spec, runtimeContext);
+  if (trustedProofs) context.proofs = trustedProofs;
+  if (policy && policy.lane === 'Direct') {
     return {
-      tier,
-      mode,
-      delegated: ['code-auditor'],
-      qualityGate: 'combined-ship-point',
-      shipPoint: mode === 'full-spec' ? 'final-task' : 'requested-task',
+      completion: 'not_applicable',
+      unfinished: false,
+      missingProof: [],
+      blocker: null,
+    };
+  }
+  return completionDecision('PASS', context);
+}
+
+function completionDecision(verdict, context = {}) {
+  if (!context || typeof context !== 'object' || Array.isArray(context)) context = {};
+  const normalized = normalizeVerdict(verdict);
+  if (normalized.verdict === 'PASS_WITH_WARNINGS') {
+    return {
+      verdict: normalized.verdict,
+      action: 'review-passed-with-warnings',
+      completion: 'unfinished',
+      terminal: true,
+      unfinished: true,
+      retry: false,
+      blocker: context.blocker || 'PASS_WITH_WARNINGS is not a completion verdict; literal PASS is required',
+    };
+  }
+  if (normalized.verdict !== 'PASS') {
+    return unfinishedVerdictDecision(normalized.verdict, context, normalized.diagnostic);
+  }
+
+  const runtimeContext = runtimeContextFromContext(context);
+  if (!runtimeContext) {
+    return {
+      verdict: normalized.verdict,
+      action: 'blocked-runtime-provenance',
+      completion: 'unfinished',
+      terminal: true,
+      unfinished: true,
+      retry: false,
+      blocker: 'completion blocked: runtime-derived provenance and session context are unavailable or stale',
+      missingProof: ['runtime_provenance'],
+    };
+  }
+  const decisionContext = { ...context, runtime_context: runtimeContext };
+
+  const missing = [];
+  const receipt = executionReceiptFromContext(decisionContext);
+  const receiptFailures = receipt
+    ? validateCanonicalReceipt(receipt, completionReceiptOptions(decisionContext))
+    : ['verification_state'];
+  if (receiptFailures.length > 0) missing.push(`execution_receipt:${receiptFailures.join(',')}`);
+
+  const obligations = workflowPolicyObligations(decisionContext);
+  if (!obligations) {
+    missing.push('workflow_policy.proof_obligations');
+  } else {
+    for (const obligation of obligations) {
+      if (obligation === 'needsExecutionProof') continue;
+      if (obligation === 'needsIndependentAudit') {
+        if (!independentAuditEvidence(decisionContext)) missing.push(obligation);
+        continue;
+      }
+      if (!evidenceForObligation(decisionContext, obligation)) missing.push(obligation);
+    }
+  }
+  if (missing.length > 0) {
+    return {
+      verdict: normalized.verdict,
+      ...(normalized.diagnostic ? { diagnostic: normalized.diagnostic } : {}),
+      action: 'await-required-proof',
+      completion: 'unfinished',
+      terminal: true,
+      unfinished: true,
+      retry: false,
+      blocker: decisionContext.blocker || `completion blocked: missing ${missing.join(', ')}`,
+      missingProof: missing,
     };
   }
   return {
-    tier,
-    mode,
-    delegated: Array.from({ length: taskCount }, () => [...DEEP_TASK_SEQUENCE]).flat(),
-    qualityGate: 'per-task-stage-a-b',
-    shipPoint: 'task',
+    verdict: normalized.verdict,
+    action: 'complete',
+    completion: 'complete',
+    terminal: true,
+    unfinished: false,
+    retry: false,
+    blocker: null,
   };
 }
 
-function consumeReviewVerdict(verdict) {
-  assertVerdict(verdict);
-  if (verdict === 'PASS') return { action: 'proceed', terminal: false };
-  if (verdict === 'FAIL') return { action: 'fix-and-rerun', terminal: false };
-  return { action: 'stop', terminal: true, blocker: 'review returned BLOCKED' };
+function consumeReviewVerdict(verdict, context = {}) {
+  const result = adaptVerdict(verdict, context);
+  return {
+    ...result,
+    source: 'review',
+    completion: 'unfinished',
+    unfinished: true,
+    createsCompletion: false,
+    completionDecisionRequired: true,
+    action: result.verdict === 'PASS' || result.verdict === 'PASS_WITH_WARNINGS'
+      ? 'review-result-only'
+      : result.action,
+    blocker: result.verdict === 'PASS' || result.verdict === 'PASS_WITH_WARNINGS'
+      ? context.blocker || 'review verdict is not completion proof'
+      : result.blocker,
+  };
 }
 
 function isFlashUnverified(task) {
   return task?.status === 'in_progress' && task.receipt === 'FLASH_UNVERIFIED';
 }
 
+function isCanonicalFlashUnverified(task) {
+  return isFlashUnverified(task)
+    && task.dependencyBlocked === true
+    && task.unblocks === false
+    && typeof task.blocker === 'string'
+    && !isPlaceholderToken(task.blocker)
+    && !FLASH_PROMOTION_FIELDS.some((field) => hasOwn(task, field));
+}
+
 function isStaleFlashDone(task) {
   return task?.status === 'done' && task.receipt === 'FLASH_UNVERIFIED';
 }
 
-function promoteFlashTask(task, verdict, proof = 'Verification: PASS') {
-  if (verdict !== 'NO_TESTS') assertVerdict(verdict);
+const SYNC_FINALIZE_CAPABILITY = Symbol('sync-finalize-capability');
+const trustedFlashPromotions = new WeakMap();
+
+function flashRuntimeContext(value) {
+  if (!PROVENANCE.isTrustedRuntimeContext(value)) return null;
+  try { return PROVENANCE.recomputeRuntimeContext(value); } catch (_) { return null; }
+}
+
+function evaluateFlashPromotion(task, verdict, proof = 'Verification: PASS', capability = null, runtimeContext) {
+  const trustedRuntime = flashRuntimeContext(runtimeContext);
+  if (!trustedRuntime) {
+    return blockedFlashSync(task, 'flash promotion blocked: runtime-derived provenance and session context are unavailable or stale');
+  }
+  const normalized = normalizeVerdict(verdict);
   if (!isFlashUnverified(task)) return task;
 
-  if (verdict !== 'PASS') {
+  if (normalized.verdict !== 'PASS') {
     return {
       ...task,
       status: 'in_progress',
       receipt: 'FLASH_UNVERIFIED',
-      blocker: verdict === 'NO_TESTS'
+      blocker: normalized.diagnostic === 'NO_TESTS'
         ? 'awaiting test proof from /hapo:test <feature>'
-        : `verification returned ${verdict}`,
+        : normalized.diagnostic === 'PARTIAL'
+          ? 'verification is partial; awaiting complete test proof'
+          : `verification returned ${normalized.verdict}`,
       dependencyBlocked: true,
       unblocks: false,
       readyForSync: false,
+      flashTransition: 'unverified',
+      promotionReceipt: null,
     };
   }
 
-  if (typeof proof !== 'string' || !/^Verification:\s*PASS\b/.test(proof.trim())) {
+  if (typeof proof !== 'string') {
     throw new TypeError('PASS promotion requires a concrete Verification: PASS receipt');
   }
+  const failures = validateCanonicalReceipt(proof, receiptValidatorOptions(task, {
+    requireProvenanceBinding: true,
+    runtimeContext: trustedRuntime,
+  }));
+  if (failures.length > 0) {
+    return {
+      ...task,
+      status: 'in_progress',
+      receipt: 'FLASH_UNVERIFIED',
+      blocker: `canonical receipt required for promotion: missing ${failures.join(', ')}`,
+      dependencyBlocked: true,
+      unblocks: false,
+      readyForSync: false,
+      flashTransition: 'unverified',
+      promotionReceipt: null,
+    };
+  }
 
-  return {
+  const promoted = {
     ...task,
     status: 'in_progress',
     receipt: proof.trim(),
@@ -391,26 +1465,110 @@ function promoteFlashTask(task, verdict, proof = 'Verification: PASS') {
     dependencyBlocked: true,
     unblocks: false,
     readyForSync: true,
+    flashTransition: 'promoted',
+    promotionReceipt: proof.trim(),
   };
+  if (capability === SYNC_FINALIZE_CAPABILITY) trustedFlashPromotions.set(promoted, capability);
+  return promoted;
 }
 
-function finalizeFlashTask(task, operation) {
-  if (operation !== 'sync-finalize') return task;
+function promoteFlashTask(task, verdict, proof = 'Verification: PASS', runtimeContext) {
+  const candidate = runtimeContext
+    || task?.runtime_context
+    || task?.runtimeContext
+    || task?.receipt_binding?.runtimeContext
+    || task?.receiptBinding?.runtimeContext;
+  return evaluateFlashPromotion(task, verdict, proof, null, candidate);
+}
+
+function finalizeFlashTaskInternal(task, operation, capability, runtimeContext) {
+  if (operation !== 'sync-finalize' || capability !== SYNC_FINALIZE_CAPABILITY || trustedFlashPromotions.get(task) !== capability) {
+    return blockedFlashSync(task, 'sync-finalize blocked: only trusted same-operation promotion may finalize');
+  }
   if (
     task?.status !== 'in_progress'
-    || !/^Verification:\s*PASS\b/.test(String(task.receipt || '').trim())
     || task.readyForSync !== true
-  ) return task;
+  ) return blockedFlashSync(task, 'sync-finalize blocked: trusted promotion state is incomplete');
+  const receipt = String(task.receipt || '');
+  if (task.flashTransition !== 'promoted' || task.promotionReceipt !== receipt) {
+    return blockedFlashSync(task, 'sync-finalize blocked: trusted promotion state is incomplete');
+  }
+  const failures = validateCanonicalReceipt(receipt, receiptValidatorOptions(task, {
+    requireProvenanceBinding: true,
+    runtimeContext,
+  }));
+  if (failures.length > 0) {
+    return blockedFlashSync(task, `sync-finalize blocked: canonical receipt required (${failures.join(', ')})`);
+  }
 
+  trustedFlashPromotions.delete(task);
   return {
     ...task,
     status: 'done',
     dependencyBlocked: false,
     unblocks: true,
     readyForSync: false,
+    flashTransition: 'finalized',
+  };
+}
+
+function finalizeFlashTask(task, operation) {
+  if (operation !== 'sync-finalize') return task;
+  return blockedFlashSync(task, 'sync-finalize blocked: only syncFinalizeFlashTask owns finalization');
+}
+
+function blockedFlashSync(task, blocker) {
+  return {
+    ...task,
+    status: 'in_progress',
+    receipt: 'FLASH_UNVERIFIED',
+    dependencyBlocked: true,
+    unblocks: false,
+    readyForSync: false,
+    flashTransition: 'unverified',
+    promotionReceipt: null,
+    blocker,
   };
 }
 
+function syncFinalizeFlashTask(task, verdict, proof, runtimeContext) {
+  if (!isCanonicalFlashUnverified(task)) {
+    return blockedFlashSync(
+      task,
+      'sync-finalize blocked: current task must be FLASH_UNVERIFIED in the exact canonical stored state; caller promotion state is not trusted',
+    );
+  }
+  if (verdict !== 'PASS') {
+    return blockedFlashSync(task, 'sync-finalize blocked: explicit PASS verdict is required');
+  }
+  if (typeof proof !== 'string' || proof.trim() === '') {
+    return blockedFlashSync(task, 'sync-finalize blocked: explicit canonical --proof is required');
+  }
+
+  // Treat all promotion fields from --task-json as untrusted. Rebuild the
+  // unverified state, then derive promotion and finalization in this operation.
+  const current = {
+    ...task,
+    readyForSync: false,
+    flashTransition: 'unverified',
+    promotionReceipt: null,
+    dependencyBlocked: true,
+    unblocks: false,
+  };
+  const candidate = runtimeContext
+    || task?.runtime_context
+    || task?.runtimeContext
+    || task?.receipt_binding?.runtimeContext
+    || task?.receiptBinding?.runtimeContext;
+  const trustedRuntime = flashRuntimeContext(candidate);
+  if (!trustedRuntime) {
+    return blockedFlashSync(task, 'sync-finalize blocked: runtime-derived provenance and session context are unavailable or stale');
+  }
+  const promoted = evaluateFlashPromotion(current, 'PASS', proof, SYNC_FINALIZE_CAPABILITY, trustedRuntime);
+  if (promoted.readyForSync !== true) return promoted;
+  return finalizeFlashTaskInternal(promoted, 'sync-finalize', SYNC_FINALIZE_CAPABILITY, trustedRuntime);
+}
+
 function flashState(taskRegistry = {}) {
   return Object.entries(taskRegistry)
     .filter(([, task]) => isFlashUnverified(task))
@@ -420,38 +1578,157 @@ function flashState(taskRegistry = {}) {
 function validateCanonicalReceipt(body, options = {}) {
   if (typeof body !== 'string') return ['verification_state'];
   const failures = [];
+  const addFailure = (failure) => {
+    if (!failures.includes(failure)) failures.push(failure);
+  };
+  // Fail-closed on placeholders anywhere — provenance, command, or artifact must not be templated
+  if (/\{\{[^}]+\}\}/.test(body)) {
+    addFailure('placeholder');
+    // placeholder also implies missing concrete provenance/command; keep explicit failures for mapping
+  }
+  // Explicit failure outcome anywhere in body must make canonical validator fail (structured only)
+  if (hasExplicitFailure(body)) {
+    addFailure('verification_state');
+  }
   // Unambiguous verification state: must be Verification: PASS exactly
   if (!/^\s*Verification:\s*PASS\s*$/m.test(body)) {
-    failures.push('verification_state');
-  }
-  // Command must be present
-  if (!/^\s*Command(?:\(s\))?\s*:/m.test(body)) {
-    failures.push('command');
-  }
-  // Exit or Result must be present with success
-  if (!/^\s*Exit\s*:|exit\s+code\s*[:=]|\bResult\s*:\s*PASS\b/im.test(body)) {
-    failures.push('exit_result');
-  }
-  // Provenance: must carry both endpoints with non-empty values on same line
-  // Base: <non-whitespace> plus Head: <non-whitespace>, OR base_sha: <non-whitespace> plus head_sha: <non-whitespace>
-  const hasBase = /^\s*Base[ \t]*:[ \t]*\S/im.test(body);
-  const hasHead = /^\s*Head[ \t]*:[ \t]*\S/im.test(body);
-  const hasBaseSha = /\bbase_sha[ \t]*:[ \t]*\S/im.test(body);
-  const hasHeadSha = /\bhead_sha[ \t]*:[ \t]*\S/im.test(body);
-  if (!((hasBase && hasHead) || (hasBaseSha && hasHeadSha))) {
-    failures.push('provenance');
-  }
-  // Artifact hash: if body mentions artifact, require sha256
-  // For strict mode, require at least one sha256 or artifact reference?
-  // Requirement says artifact or evidence hash when applicable. We make it optional but if artifact mentioned, require hash.
-  // For done tasks that produce artifacts, hash should be present. We enforce that if body contains Artifact|artifact, then sha256 must be present.
-  if (/\bartifact\b/i.test(body) && !/sha256:/i.test(body)) {
-    failures.push('artifact_hash');
-  }
-  // Evidence hash: if no artifact but strict evidence hash required, we could require sha256 generally for done tasks?
-  // Make it optional for now; but we can require at least one hash when requireArtifactHash true
-  if (options.requireArtifactHash && !/sha256:/i.test(body)) {
-    failures.push('artifact_hash');
+    addFailure('verification_state');
+  }
+  // Command must be present with non-empty concrete value on same line, not placeholder
+  const cmdLine = body.split('\n').find((l) => /^\s*Command(?:\(s\))?\s*:/m.test(l)) || null;
+  if (!cmdLine || !/^\s*Command(?:\(s\))?\s*:[ \t]*\S/m.test(cmdLine)) {
+    addFailure('command');
+  } else {
+    const m = cmdLine.match(/^\s*Command(?:\(s\))?\s*:[ \t]*(.*)$/m);
+    const val = m ? m[1].trim() : '';
+    if (isPlaceholderToken(val)) {
+      addFailure('command');
+    } else if (/\{\{[^}]+\}\}/.test(cmdLine)) {
+      addFailure('command');
+    }
+  }
+  // Exit / Result handling: collect all Result lines; if no Exit then need at least one Result and all PASS; if has Exit then every Exit integer 0 and every Result if any all PASS
+  const lines = body.split('\n');
+  const exitValues = [];
+  const resultValues = [];
+  for (const line of lines) {
+    if (/^\s*Exit\s*:/i.test(line)) {
+      const m = line.match(/^\s*Exit\s*:\s*(.*)$/i);
+      exitValues.push(m ? m[1].trim() : '');
+    } else if (/exit\s+code\s*[:=]/i.test(line)) {
+      const m = line.match(/exit\s+code\s*[:=]\s*(.*)$/i);
+      if (m) exitValues.push(m[1].trim());
+    }
+    if (/^\s*Result\s*:/i.test(line)) {
+      const m = line.match(/^\s*Result\s*:\s*(.+?)\s*$/i);
+      resultValues.push(m ? m[1].trim() : '');
+    }
+  }
+
+  if (exitValues.length === 0) {
+    if (resultValues.length === 0 || resultValues.some((v) => v !== 'PASS')) {
+      addFailure('exit_result');
+    }
+  } else {
+    let exitOk = true;
+    for (const v of exitValues) {
+      if (!/^-?\d+$/.test(v)) { exitOk = false; break; }
+      const n = Number.parseInt(v, 10);
+      if (n !== 0) { exitOk = false; break; }
+    }
+    if (!exitOk) addFailure('exit_result');
+    if (resultValues.length > 0 && resultValues.some((v) => v !== 'PASS')) {
+      addFailure('exit_result');
+    }
+  }
+  // Provenance is either a concrete Base/Head pair or a concrete base_sha/head_sha pair.
+  // Without expected values this validates schema only; it does not claim identity binding.
+  const receiptLines = body.split('\n');
+  const readField = (name) => receiptLines
+    .filter((line) => new RegExp(`^\\s*${name}\\s*:`, 'i').test(line))
+    .map((line) => line.replace(new RegExp(`^\\s*${name}\\s*:\\s*`, 'i'), '').trim());
+  const baseValues = readField('Base');
+  const headValues = readField('Head');
+  const baseShaValues = readField('base_sha');
+  const headShaValues = readField('head_sha');
+  const labelStyle = baseValues.length > 0 || headValues.length > 0;
+  const shaStyle = baseShaValues.length > 0 || headShaValues.length > 0;
+  let actualProvenance = null;
+  if (labelStyle === shaStyle || labelStyle && (baseShaValues.length > 0 || headShaValues.length > 0)) {
+    addFailure('provenance');
+  } else {
+    const base = labelStyle ? baseValues : baseShaValues;
+    const head = labelStyle ? headValues : headShaValues;
+    if (base.length !== 1 || head.length !== 1 || !isConcreteProvenanceValue(base[0]) || !isConcreteProvenanceValue(head[0])) {
+      addFailure('provenance');
+    } else {
+      actualProvenance = { base: normalizeProvenanceValue(base[0]), head: normalizeProvenanceValue(head[0]) };
+    }
+  }
+  const expected = options.expectedProvenance || options.expected_provenance || options.expected || null;
+  const expectedBase = expected && typeof expected === 'object'
+    ? firstDefined(expected, ['base', 'Base', 'base_sha', 'baseSha', 'expectedBase', 'expected_base'])
+    : options.expectedBase ?? options.expected_base ?? options.expectedBaseSha ?? options.expected_base_sha;
+  const expectedHead = expected && typeof expected === 'object'
+    ? firstDefined(expected, ['head', 'Head', 'head_sha', 'headSha', 'expectedHead', 'expected_head'])
+    : options.expectedHead ?? options.expected_head ?? options.expectedHeadSha ?? options.expected_head_sha;
+  if (options.requireProvenanceBinding === true || expectedBase !== undefined || expectedHead !== undefined) {
+    if (!actualProvenance
+      || !isConcreteProvenanceValue(expectedBase)
+      || !isConcreteProvenanceValue(expectedHead)
+      || normalizeProvenanceValue(expectedBase) !== actualProvenance.base
+      || normalizeProvenanceValue(expectedHead) !== actualProvenance.head) {
+      addFailure('provenance');
+    }
+  }
+
+  if (options.artifactDeclarationValid === false) addFailure('artifact_declaration');
+  const artifactDeclarations = parseArtifactDeclarations(body);
+  for (const declaration of artifactDeclarations) {
+    if (declaration.hashes.length === 0 || declaration.hashes.some((hash) => !/^[a-f0-9]{64}$/i.test(hash))) {
+      addFailure('artifact_hash');
+    }
+    const declaredPaths = declaration.pathText
+      .split(/\s*(?:\+|,|;|\band\b)\s*/i)
+      .map((value) => value.trim())
+      .filter(Boolean);
+    if (declaredPaths.length > 1 && declaration.hashes.length !== declaredPaths.length) {
+      addFailure('artifact_hash');
+    }
+  }
+  const artifactPaths = Array.isArray(options.artifactPaths) ? options.artifactPaths : [];
+  if (options.requireArtifactHash) {
+    if (artifactDeclarations.length === 0) addFailure('artifact_hash');
+    for (const artifactPath of artifactPaths) {
+      const matches = artifactDeclarations.filter((declaration) => {
+        const candidates = declaration.pathText.split(/\s*(?:\+|,|;|\band\b)\s*/i).map((value) => value.trim()).filter(Boolean);
+        return candidates.length === 1 && candidates.includes(artifactPath);
+      });
+      if (matches.length !== 1 || matches[0].hashes.length !== 1 || !/^[a-f0-9]{64}$/i.test(matches[0].hashes[0])) {
+        addFailure('artifact_hash');
+      }
+    }
+  }
+  if (options.verifyArtifactBytes === true || typeof options.artifactRoot === 'string') {
+    if (typeof options.artifactRoot !== 'string' || options.artifactRoot.trim() === '') {
+      if (artifactDeclarations.length > 0 || options.requireArtifactHash) addFailure('artifact_hash');
+    } else {
+      for (const declaration of artifactDeclarations) {
+        const declaredPaths = declaration.pathText
+          .split(/\s*(?:\+|,|;|\band\b)\s*/i)
+          .map((value) => value.trim())
+          .filter(Boolean);
+        if (declaredPaths.length === 0 || declaration.hashes.length !== declaredPaths.length) {
+          addFailure('artifact_hash');
+          continue;
+        }
+        for (let index = 0; index < declaredPaths.length; index += 1) {
+          if (!hashValidatedArtifact(options.artifactRoot, declaredPaths[index], declaration.hashes[index])) {
+            addFailure('artifact_hash');
+          }
+        }
+      }
+    }
   }
   return failures;
 }
@@ -466,9 +1743,16 @@ function parseCliArgs(argv) {
     action: null,
     lane: null,
     override: null,
-    proof: 'Verification: PASS',
+    proof: null,
     userAuthorized: false,
     origin: null,
+    projectRoot: null,
+    specsRoot: null,
+    specFile: null,
+    featureName: null,
+    runtimeSession: null,
+    provenanceMode: null,
+    runtimeContextInput: undefined,
   };
   const args = [...argv];
   for (let i = 0; i < args.length; i += 1) {
@@ -483,6 +1767,9 @@ function parseCliArgs(argv) {
     else if (arg === '--sync-finalize') options.action = 'sync-finalize';
     else if (arg === '--classify-lane') options.action = 'classify-lane';
     else if (arg === '--lane-policy') options.action = 'lane-policy';
+    else if (arg === '--policy-snapshot') options.action = 'policy-snapshot';
+    else if (arg === '--validate-policy') options.action = 'validate-policy';
+    else if (arg === '--completion-decision') options.action = 'completion-decision';
     else if (arg === '--approval-state') options.action = 'approval-state';
     else if (arg === '--validate-approval') options.action = 'validate-approval';
     else if (arg === '--validate-receipt') options.action = 'validate-receipt';
@@ -498,6 +1785,16 @@ function parseCliArgs(argv) {
     } else if (arg === '--proof') {
       options.proof = args[++i];
       if (!options.proof) throw new Error('--proof requires a value');
+    } else if (arg === '--project-root') options.projectRoot = args[++i];
+    else if (arg === '--specs-root') options.specsRoot = args[++i];
+    else if (arg === '--spec-file') options.specFile = args[++i];
+    else if (arg === '--feature-name') options.featureName = args[++i];
+    else if (arg === '--runtime-session' || arg === '--session') options.runtimeSession = args[++i];
+    else if (arg === '--provenance-mode') options.provenanceMode = args[++i];
+    else if (arg === '--runtime-context') {
+      const raw = args[++i];
+      if (!raw) throw new Error('--runtime-context requires a JSON value');
+      options.runtimeContextInput = JSON.parse(raw);
     } else {
       throw new Error(`Unknown option: ${arg}`);
     }
@@ -505,6 +1802,23 @@ function parseCliArgs(argv) {
   return options;
 }
 
+function cliRuntimeContext(options) {
+  if (options.runtimeContextInput !== undefined) {
+    throw new Error('--runtime-context JSON is not trusted; provide runtime identity flags');
+  }
+  const supplied = [options.projectRoot, options.specsRoot, options.specFile, options.featureName, options.runtimeSession]
+    .some((value) => value !== null && value !== undefined);
+  if (!supplied) return null;
+  return PROVENANCE.deriveRuntimeContext({
+    projectRoot: options.projectRoot,
+    specsRoot: options.specsRoot,
+    specFile: options.specFile,
+    featureName: options.featureName,
+    runtimeSession: options.runtimeSession,
+    provenanceMode: options.provenanceMode || undefined,
+  });
+}
+
 function cliResult(result, json) {
   process.stdout.write(json ? `${JSON.stringify(result)}\n` : `${result.message}\n`);
   return result.exitCode;
@@ -541,6 +1855,17 @@ function runCli(argv = process.argv.slice(2)) {
       const policy = lanePolicy(input);
       return cliResult({ ok: true, policy, exitCode: 0, message: `Lane: ${policy.lane}` }, options.json);
     }
+    if (options.action === 'policy-snapshot') {
+      if (!options.task) throw new Error('--policy-snapshot requires --task-json');
+      const snapshot = workflowPolicySnapshot(options.task);
+      return cliResult({ ok: true, snapshot, exitCode: 0, message: `Policy snapshot: ${snapshot.lane}` }, options.json);
+    }
+    if (options.action === 'validate-policy') {
+      if (!options.task) throw new Error('--validate-policy requires --task-json');
+      const snapshot = hasOwn(options.task, 'workflow_policy') ? options.task.workflow_policy : options.task;
+      const result = validateWorkflowPolicySnapshot(snapshot);
+      return cliResult({ ok: result.valid, ...result, exitCode: result.valid ? 0 : 2, message: result.valid ? 'Workflow policy snapshot valid' : result.errors.join('; ') }, options.json);
+    }
     if (options.action === 'approval-state') {
       if (!options.task) throw new Error('--approval-state requires --task-json');
       const state = approvalState(options.task);
@@ -554,22 +1879,30 @@ function runCli(argv = process.argv.slice(2)) {
     if (options.action === 'validate-receipt') {
       if (!options.task) throw new Error('--validate-receipt requires --task-json');
       const body = options.task.body || options.task.receipt || '';
-      const failures = validateCanonicalReceipt(body);
+      const failures = validateCanonicalReceipt(body, receiptValidatorOptions(options.task, { runtimeContext: cliRuntimeContext(options) }));
       return cliResult({ ok: failures.length === 0, failures, exitCode: failures.length === 0 ? 0 : 2, message: failures.length === 0 ? 'Receipt valid' : `Receipt missing: ${failures.join(', ')}` }, options.json);
     }
     if (options.action === 'consume-verdict') {
-      const result = consumeReviewVerdict(options.verdict);
+      const result = consumeReviewVerdict(options.verdict, options.task || {});
       return cliResult({ ok: true, ...result, exitCode: 0, message: `Verdict ${options.verdict}: ${result.action}` }, options.json);
     }
+    if (options.action === 'completion-decision') {
+      const runtimeContext = cliRuntimeContext(options);
+      const result = completionDecision(options.verdict, runtimeContext
+        ? { ...(options.task || {}), runtime_context: runtimeContext }
+        : (options.task || {}));
+      return cliResult({ ok: true, ...result, exitCode: 0, message: `Verdict ${options.verdict}: ${result.completion}` }, options.json);
+    }
     if (options.action === 'promote-flash') {
       if (!options.task) throw new Error('--promote-flash requires --task-json');
-      const result = promoteFlashTask(options.task, options.verdict, options.proof);
+      const result = promoteFlashTask(options.task, options.verdict, options.proof, cliRuntimeContext(options));
       return cliResult({ ok: true, task: result, exitCode: 0, message: 'Flash task promotion evaluated.' }, options.json);
     }
     if (options.action === 'sync-finalize') {
       if (!options.task) throw new Error('--sync-finalize requires --task-json');
-      const result = finalizeFlashTask(options.task, 'sync-finalize');
-      return cliResult({ ok: true, task: result, exitCode: 0, message: 'Flash task sync-finalize evaluated.' }, options.json);
+      const result = syncFinalizeFlashTask(options.task, options.verdict, options.proof, cliRuntimeContext(options));
+      const ok = result.status === 'done';
+      return cliResult({ ok, task: result, exitCode: ok ? 0 : 2, message: ok ? 'Flash task sync-finalize evaluated.' : (result.blocker || 'Flash task sync-finalize blocked.') }, options.json);
     }
 
     return cliResult({ ok: true, contract: 'execution-policy', ...policy, exitCode: 0, message: `Execution mode: ${policy.mode}` }, options.json);
@@ -583,26 +1916,60 @@ if (require.main === module) process.exitCode = runCli();
 
 module.exports = {
   REVIEW_VERDICTS,
+  CANONICAL_VERDICTS,
+  LEGACY_DIAGNOSTIC_VERDICTS,
   EXECUTION_TIERS,
   LANES,
+  ARTIFACT_PROFILES,
+  PROOF_OBLIGATIONS,
+  WORKFLOW_POLICY_VERSION,
+  WORKFLOW_POLICY_FIELDS,
+  RECEIPT_BINDING_FIELDS,
+  INDEPENDENT_AUDIT_SCHEMA_VERSION,
+  INDEPENDENT_AUDIT_FIELDS,
   APPROVAL_SCHEMA_VERSION,
   LEGACY_APPROVAL_ERROR,
-  DEEP_TASK_SEQUENCE,
-  CRITICAL_LANE_SEQUENCE,
   executionPolicy,
   classifyLane,
+  obligationsForLane,
+  actorNeedsFor,
+  isWorkflowPolicySnapshot,
+  validateWorkflowPolicySnapshot,
+  assertWorkflowPolicySnapshot,
+  workflowPolicySnapshot,
+  createWorkflowPolicySnapshot,
+  persistWorkflowPolicySnapshot,
+  persistWorkflowPolicy,
+  readWorkflowPolicySnapshot,
+  escalateWorkflowPolicy,
+  policyForSpec,
   lanePolicy,
   approvalState,
   validateApprovalSchema,
   isLegacyApprovalObject,
   isUserAuthorizedForDowngrade,
+  isValidOverrideReceipt,
+  isTapMetadataHeading,
+  deriveRuntimeContext: PROVENANCE.deriveRuntimeContext,
+  recomputeRuntimeContext: PROVENANCE.recomputeRuntimeContext,
+  isTrustedRuntimeContext: PROVENANCE.isTrustedRuntimeContext,
+  createReceiptBinding,
+  receiptValidatorOptions,
   validateCanonicalReceipt,
+  validateIndependentAuditEvidence,
+  completionDecisionForSpec,
   delegationPlan,
+  assertVerdict,
+  normalizeVerdict,
+  adaptVerdict,
+  completionDecision,
   consumeReviewVerdict,
   isFlashUnverified,
+  isCanonicalFlashUnverified,
   isStaleFlashDone,
   promoteFlashTask,
   finalizeFlashTask,
+  syncFinalizeFlashTask,
   flashState,
   parseCliArgs,
   runCli,