peaks-loop 3.1.2 → 4.0.0-beta.10

package/dist/src/services/workflow/pipeline-verify-service.d.ts

@@ -1,90 +0,0 @@
-import type { RequestType } from '../artifacts/artifact-prerequisites.js';
-export type PipelineGate = {
-    name: string;
-    description: string;
-    passed: boolean;
-    detail: string;
-    /**
-     * Slice 2026-06-13-peaks-workflow-skip: optional status that
-     * distinguishes a gate that was BYPASSED via `peaks workflow skip`
-     * from one that actually passed evaluation. When set, the gate
-     * was NOT evaluated by `evaluateGate` — the user/CLI explicitly
-     * marked it as skipped. The boolean `passed` is set to `true` so
-     * existing consumers (which only check `passed`) treat a skipped
-     * gate as "satisfied", but downstream consumers that need
-     * audit-grade distinction (e.g. CI badges, dashboards) can read
-     * `status` to tell pass / fail / skipped apart.
-     *
-     *   - omitted (default): not set, gate was evaluated; treat as legacy.
-     *   - 'pass':          gate evaluated and passed.
-     *   - 'fail':          gate evaluated and failed.
-     *   - 'skipped':       gate was bypassed; never evaluated.
-     */
-    status?: 'pass' | 'fail' | 'skipped';
-};
-export type PipelineVerification = {
-    rid: string;
-    sessionId: string;
-    requestType: RequestType;
-    complete: boolean;
-    rdPhase: {
-        invoked: boolean;
-        state: string;
-        gates: PipelineGate[];
-    };
-    qaPhase: {
-        invoked: boolean;
-        state: string;
-        gates: PipelineGate[];
-    };
-    /**
-     * Slice 002 (v2.15.0) AC-3: Gate H "feedback-promotion". Always
-     * present (single-element array). Evaluates whether every
-     * `metadata.type === 'feedback'` memory in `.peaks/memory/`
-     * carries a promotion marker (comment OR sidecar). Failures
-     * block the `complete` verdict via the `gateH` field below; the
-     * pipeline only completes when every gate in this array passes.
-     */
-    feedbackPhase?: {
-        gates: PipelineGate[];
-    };
-    violations: string[];
-    nextActions: string[];
-    /** Form of the security/performance findings artifacts Gate C accepted
-     * (slice 025). `'suffixed'` for the new per-rid form, `'legacy'` for the
-     * pre-slice-025 non-suffixed form, `'none'` when neither was found. */
-    acceptedForm?: 'suffixed' | 'legacy' | 'none';
-    /** `gateC` is the pre-computed verdict string (AC7 dogfood shape). */
-    gateC?: 'pass' | 'fail';
-    /**
-     * Slice 002 (v2.15.0) AC-3: Gate H verdict string. `'pass'` when
-     * all feedback memories are promoted; `'fail'` when at least one
-     * unpromoted feedback memory was found.
-     */
-    gateH?: 'pass' | 'fail';
-    /**
-     * Slice 2026-06-28-solo-mode-bypass-fix (defect #3): `true` when
-     * every evidence file resolved on the canonical path
-     * (`.peaks/_runtime/change/<sessionId>/...`). `false` when at least
-     * one evidence file resolved via a legacy fallback (`.peaks/<sessionId>/...`
-     * or `.peaks/_runtime/<sessionId>/...`). QA / TXT surface the value
-     * so users know to move misplaced evidence into the canonical
-     * location.
-     *
-     * Slice 2026-06-29-change-id-root-removal: the legacy
-     * `peaks workspace migrate-change-scope` migration tool is gone;
-     * users must now move misplaced content into the canonical
-     * `.peaks/_runtime/change/<sessionId>/<role>/` dir manually (or via
-     * `peaks workspace migrate`).
-     */
-    usedCanonicalPath?: boolean;
-};
-export declare function verifyPipeline(options: {
-    projectRoot: string;
-    rid: string;
-    /** Optional explicit session-id; when omitted, the RD/QA on-disk location
-     * is resolved via showRequestArtifact (which scans all top-level dirs and
-     * returns the actual session-id the file lives in). */
-    sessionId?: string;
-    requestType?: string;
-}): Promise<PipelineVerification>;

package/dist/src/services/workflow/pipeline-verify-service.js

@@ -1,449 +0,0 @@
-import { existsSync } from 'node:fs';
-import { join } from 'node:path';
-import { isRequestType } from '../artifacts/artifact-prerequisites.js';
-import { showRequestArtifact } from '../artifacts/request-artifact-service.js';
-import { resolveSecurityFindingsPath, resolvePerformanceFindingsPath } from './artifact-paths.js';
-import { readSkipState } from './workflow-state-store.js';
-import { getSessionIdCanonical } from '../session/session-manager.js';
-import { listUnpromotedFeedback } from '../feedback/feedback-promotion-service.js';
-function extractState(markdown) {
-    for (const rawLine of markdown.split(/\r?\n/)) {
-        const match = /^-\s*state:\s*(.+?)\s*$/.exec(rawLine.trim());
-        if (match?.[1])
-            return match[1];
-    }
-    return 'unknown';
-}
-/**
- * As of slice 2026-06-05-change-id-as-unit-of-work, the file's durable
- * scope is the change-id (the `.peaks/_runtime/<sessionId>/` dir the file lives
- * in), NOT the session-id. We resolve the on-disk location via
- * `showRequestArtifact` (which scans all top-level dirs and returns the
- * actual dir the file was found in) instead of assuming
- * `.peaks/_runtime/<sessionId>/<role>/requests/`.
- */
-async function findRequestFile(projectRoot, role, rid) {
-    const artifact = await showRequestArtifact({ projectRoot, role: role, requestId: rid });
-    if (artifact === null)
-        return null;
-    // Slice 2026-06-28-solo-mode-bypass-fix (defect #3): the legacy
-    // `showRequestArtifact` returns the FULL SCOPE (`_runtime/<sid>`)
-    // as `sessionId`, not just the trailing id segment. The canonical
-    // evidence lookup needs only the bare id (`.peaks/_runtime/change/<id>/`).
-    // When the scope starts with `_runtime/`, strip that prefix so the
-    // path resolver builds the right canonical location.
-    let sessionId = artifact.sessionId;
-    if (sessionId.startsWith('_runtime/') || sessionId.startsWith('_runtime\\')) {
-        sessionId = sessionId.replace(/^_runtime[\\/]/, '');
-    }
-    return { path: artifact.path, content: artifact.content, sessionId };
-}
-function rdGatesForType(requestType) {
-    const gates = [
-        { name: 'rd-request-exists', description: 'RD request artifact created', passed: false, detail: '' }
-    ];
-    if (requestType === 'feature' || requestType === 'refactor') {
-        gates.push({ name: 'tech-doc', description: 'Technical design doc', passed: false, detail: '' });
-    }
-    if (requestType === 'bugfix') {
-        gates.push({ name: 'bug-analysis', description: 'Bug root-cause analysis', passed: false, detail: '' });
-    }
-    if (requestType !== 'docs' && requestType !== 'chore' && requestType !== 'config') {
-        gates.push({ name: 'code-review', description: 'Code review evidence', passed: false, detail: '' });
-    }
-    if (requestType === 'feature' || requestType === 'refactor' || requestType === 'bugfix' || requestType === 'config') {
-        gates.push({ name: 'security-review', description: 'Security review evidence', passed: false, detail: '' });
-    }
-    return gates;
-}
-function qaGatesForType(requestType) {
-    const gates = [
-        { name: 'qa-request-exists', description: 'QA request artifact created', passed: false, detail: '' }
-    ];
-    if (requestType === 'feature' || requestType === 'refactor' || requestType === 'bugfix') {
-        gates.push({ name: 'test-cases', description: 'QA test cases', passed: false, detail: '' });
-        gates.push({ name: 'test-report', description: 'QA test report with execution results', passed: false, detail: '' });
-    }
-    if (requestType === 'feature' || requestType === 'refactor' || requestType === 'bugfix' || requestType === 'config') {
-        gates.push({ name: 'security-findings', description: 'QA security findings', passed: false, detail: '' });
-    }
-    if (requestType === 'feature' || requestType === 'refactor') {
-        gates.push({ name: 'performance-findings', description: 'QA performance findings', passed: false, detail: '' });
-    }
-    return gates;
-}
-const RD_QA_HANDOFF_STATES = new Set(['qa-handoff', 'handed-off', 'implemented']);
-const QA_COMPLETE_STATES = new Set(['verdict-issued']);
-export async function verifyPipeline(options) {
-    const requestType = isRequestType(options.requestType ?? '') ? options.requestType : 'feature';
-    const violations = [];
-    const nextActions = [];
-    const rdGates = rdGatesForType(requestType);
-    const qaGates = qaGatesForType(requestType);
-    // Slice 2026-06-13-peaks-workflow-skip: read the skip-state (if any)
-    // and pre-mark matching gates as `status: 'skipped'`. We do this
-    // BEFORE evaluating evidence files, so a skipped gate never
-    // emits a "missing evidence" violation. The boolean `passed` is
-    // set to `true` so legacy consumers (which only check `passed`)
-    // treat skipped gates as satisfied; the new `status` field
-    // distinguishes the bypass from an actual pass.
-    const skippedGateNames = new Set();
-    if (options.sessionId !== undefined) {
-        const skipState = readSkipState(options.projectRoot, options.sessionId, options.rid);
-        if (skipState !== null) {
-            for (const gateName of skipState.skippedGates) {
-                if (gateName === 'QA' || gateName === 'qa-phase' || gateName === 'qa') {
-                    for (const g of qaGates)
-                        skippedGateNames.add(g.name);
-                }
-                else if (gateName === 'RD' || gateName === 'rd-phase' || gateName === 'rd') {
-                    for (const g of rdGates)
-                        skippedGateNames.add(g.name);
-                }
-                else {
-                    skippedGateNames.add(gateName);
-                }
-            }
-            nextActions.push(`Skipped gates (via peaks workflow skip): [${skipState.skippedGates.join(', ')}] — reason: ${skipState.skipReason}`);
-        }
-    }
-    function markIfSkipped(gate) {
-        if (skippedGateNames.has(gate.name)) {
-            gate.status = 'skipped';
-            gate.passed = true;
-            gate.detail = 'skipped via peaks workflow skip';
-        }
-    }
-    // Resolve RD + QA on-disk locations via showRequestArtifact (the change-id
-    // is whatever dir the file actually lives in, not the caller's session-id).
-    const rdFile = await findRequestFile(options.projectRoot, 'rd', options.rid);
-    let rdInvoked = false;
-    let rdState = 'missing';
-    // The resolved change-id is the on-disk location the file actually
-    // lives in. The caller's `options.sessionId` is a hint used for
-    // path construction (nextActions strings), NOT for the resolved
-    // sessionId field — the on-disk location is the source of truth.
-    let resolvedChangeId = '';
-    if (rdFile) {
-        rdInvoked = true;
-        rdState = extractState(rdFile.content);
-        rdGates[0].passed = true;
-        rdGates[0].detail = `found at ${rdFile.path}`;
-        resolvedChangeId = rdFile.sessionId;
-    }
-    else {
-        violations.push('RD phase skipped: peaks-rd was never invoked for this request (no RD request artifact found)');
-        nextActions.push('Invoke Skill(skill="peaks-rd") with the request-id, then run unit tests + code review + security review');
-        rdGates[0].detail = 'not found';
-    }
-    // Check RD evidence files (under the change-id dir the RD request lives in)
-    const RD_EVIDENCE_FILE = {
-        'tech-doc': 'tech-doc.md',
-        'bug-analysis': 'bug-analysis.md',
-        'code-review': 'code-review.md',
-        'security-review': 'security-review.md'
-    };
-    // Slice v2.17.0 hard-kill — change-id axis removed as filesystem
-    // scope. The canonical evidence location is now the session axis
-    // `.peaks/_runtime/<sessionId>/<role>/...` (where `<sessionId>` is
-    // the on-disk dir the request artifact lives in; for back-compat
-    // with the legacy `_runtime/<sessionId>/` and `.peaks/<sessionId>/`
-    // layouts we also probe those). When the legacy fallback fires, the
-    // gate detail / nextActions surface the `DEPRECATION_LEGACY_PATH_USED`
-    // warning so QA / TXT can nudge users to migrate via
-    // `peaks workspace migrate-change-scope --apply`.
-    // v2.18.1 bug #5 fix: same session-axis fallback as the QA resolver
-    // below. When `resolvedChangeId` is empty (no RD/QA artifact on disk
-    // yet), the canonical session id from the binding-store is the
-    // preferred filesystem scope; falling through to `options.rid` would
-    // make every missing-evidence path look like a per-rid scope dir.
-    const rdEvidenceDir = resolvedChangeId || options.sessionId || getSessionIdCanonical(options.projectRoot) || options.rid;
-    // Slice 2026-06-28-solo-mode-bypass-fix (defect #3): track whether
-    // every resolved evidence file landed on the canonical path. Drives
-    // the `usedCanonicalPath` envelope field. The flag stays `true` if
-    // no evidence file was resolved at all (nothing to be non-canonical
-    // about); it flips to `false` if any file resolves via legacy.
-    let anyEvidenceResolved = false;
-    let allResolvedPathsCanonical = true;
-    for (const gate of rdGates.slice(1)) {
-        const fileName = RD_EVIDENCE_FILE[gate.name];
-        // v2.17.0 canonical: session axis — `<sessionId>` is the on-disk
-        // dir the request artifact was found in (matches the layout used
-        // by `request-artifact-service.ts` for new artifacts).
-        const canonicalPath = join(options.projectRoot, '.peaks', '_runtime', rdEvidenceDir, 'rd', fileName);
-        const legacyMisplacedPath = join(options.projectRoot, '.peaks', rdEvidenceDir, 'rd', fileName);
-        // Pre-v2.17.0 change-axis form: `.peaks/_runtime/change/<id>/rd/<file>`
-        // — the v2.16.0 / v2.17.0-era canonical that v2.17.0 hard-killed.
-        // Kept as a back-compat fallback for un-migrated workspaces.
-        const legacyChangeAxisPath = join(options.projectRoot, '.peaks', '_runtime', 'change', rdEvidenceDir, 'rd', fileName);
-        let resolvedPath = null;
-        let usedLegacy = false;
-        for (const candidate of [canonicalPath, legacyMisplacedPath, legacyChangeAxisPath]) {
-            if (existsSync(candidate)) {
-                resolvedPath = candidate;
-                usedLegacy = candidate !== canonicalPath;
-                break;
-            }
-        }
-        if (resolvedPath !== null) {
-            anyEvidenceResolved = true;
-            if (usedLegacy)
-                allResolvedPathsCanonical = false;
-            gate.passed = true;
-            gate.detail = resolvedPath + (usedLegacy ? ' [DEPRECATION_LEGACY_PATH_USED]' : '');
-            if (usedLegacy) {
-                violations.push(`DEPRECATION_LEGACY_PATH_USED: ${resolvedPath} — move the file into .peaks/_runtime/${rdEvidenceDir}/rd/ (the canonical location) so subsequent runs resolve on the canonical path. The legacy \`peaks workspace migrate-change-scope\` helper was removed in v2.19.0; use \`peaks workspace migrate\` to relocate misplaced content.`);
-            }
-        }
-        else {
-            gate.detail = `missing: ${canonicalPath}`;
-            violations.push(`RD evidence missing: ${gate.description} (${fileName})`);
-            nextActions.push(`Create .peaks/_runtime/${rdEvidenceDir}/rd/${fileName}`);
-        }
-    }
-    // Check if RD reached qa-handoff
-    if (rdInvoked && !RD_QA_HANDOFF_STATES.has(rdState)) {
-        violations.push(`RD not ready for QA: state is "${rdState}" — must reach "qa-handoff" (unit tests, karpathy-guidelines §1 Think / §2 Simplicity / §3 Surgical / §4 Goal-Driven, code review, security review complete)`);
-        nextActions.push(`Complete RD gates → peaks request transition ${options.rid} --role rd --state qa-handoff`);
-    }
-    // Check QA phase
-    const qaFile = await findRequestFile(options.projectRoot, 'qa', options.rid);
-    let qaInvoked = false;
-    let qaState = 'missing';
-    if (qaFile) {
-        qaInvoked = true;
-        qaState = extractState(qaFile.content);
-        qaGates[0].passed = true;
-        qaGates[0].detail = `found at ${qaFile.path}`;
-        resolvedChangeId = qaFile.sessionId || resolvedChangeId;
-    }
-    else {
-        violations.push('QA phase skipped: peaks-qa was never invoked for this request (no QA request artifact found)');
-        nextActions.push('Invoke Skill(skill="peaks-qa") with the request-id for functional/performance/security testing');
-        qaGates[0].detail = 'not found';
-    }
-    // Check QA evidence files. For the security/performance findings
-    // gates, the canonical form post-slice-025 is the per-rid suffixed
-    // path; the legacy non-suffixed form is still accepted during the
-    // 1-minor-release back-compat window. The path resolver
-    // (artifact-paths.ts) decides which form to consume; we pass the
-    // resolved change-id and the rid.
-    // v2.18.1 bug #5 fix: when no RD/QA artifact is on disk yet (resolvedChangeId
-    // is empty), fall back to the current session id from the binding-store
-    // instead of `rdEvidenceDir` (= the rid). The session axis
-    // `.peaks/_runtime/<sessionId>/qa/...` is the canonical v2.17.0 home; the
-    // legacy `_runtime/change/<sessionId>/qa/...` probe should only fire for
-    // pre-v2.17.0 workspaces, not as a default for new requests. The slug in
-    // the `sessionId` envelope field is preserved below for traceability.
-    const changeIdForResolver = resolvedChangeId || getSessionIdCanonical(options.projectRoot) || rdEvidenceDir;
-    const QA_EVIDENCE_FILE = {
-        'test-cases': `test-cases/${options.rid}.md`,
-        'test-report': `test-reports/${options.rid}.md`,
-        'security-findings': '',
-        'performance-findings': ''
-    };
-    for (const gate of qaGates.slice(1)) {
-        if (gate.name === 'security-findings' || gate.name === 'performance-findings') {
-            const resolver = gate.name === 'security-findings' ? resolveSecurityFindingsPath : resolvePerformanceFindingsPath;
-            const resolved = resolver({ projectRoot: options.projectRoot, sessionId: changeIdForResolver, rid: options.rid });
-            if (existsSync(resolved.path)) {
-                anyEvidenceResolved = true;
-                if (resolved.form === 'legacy')
-                    allResolvedPathsCanonical = false;
-                gate.passed = true;
-                gate.detail = resolved.path;
-                if (resolved.form === 'legacy') {
-                    // 1-minor-release back-compat window. Surface the warning so
-                    // users know to migrate. Per PRD §Migration the form will be
-                    // rejected after the next minor bump.
-                    violations.push(`QA evidence accepted in legacy form (will be rejected after next minor release): ${resolved.path} — re-run peaks workflow plan refresh to migrate`);
-                }
-            }
-            else {
-                gate.detail = `missing: ${resolved.path}`;
-                violations.push(`QA evidence missing: ${gate.description} (${resolved.path})`);
-                nextActions.push(`Create ${resolved.path} (or use the legacy non-suffixed form during the 1-minor-release back-compat window)`);
-            }
-            continue;
-        }
-        const fileName = QA_EVIDENCE_FILE[gate.name];
-        // v2.17.0 canonical: session axis — `<sessionId>/qa/<file>`.
-        const canonicalQaPath = join(options.projectRoot, '.peaks', '_runtime', rdEvidenceDir, 'qa', fileName);
-        const legacyMisplacedQaPath = join(options.projectRoot, '.peaks', rdEvidenceDir, 'qa', fileName);
-        // Pre-v2.17.0 change-axis form: kept as back-compat fallback.
-        const legacyChangeAxisQaPath = join(options.projectRoot, '.peaks', '_runtime', 'change', rdEvidenceDir, 'qa', fileName);
-        let resolvedQaPath = null;
-        let usedLegacyQa = false;
-        for (const candidate of [canonicalQaPath, legacyMisplacedQaPath, legacyChangeAxisQaPath]) {
-            if (existsSync(candidate)) {
-                resolvedQaPath = candidate;
-                usedLegacyQa = candidate !== canonicalQaPath;
-                break;
-            }
-        }
-        if (resolvedQaPath !== null) {
-            anyEvidenceResolved = true;
-            if (usedLegacyQa)
-                allResolvedPathsCanonical = false;
-            gate.passed = true;
-            gate.detail = resolvedQaPath + (usedLegacyQa ? ' [DEPRECATION_LEGACY_PATH_USED]' : '');
-            if (usedLegacyQa) {
-                violations.push(`DEPRECATION_LEGACY_PATH_USED: ${resolvedQaPath} — move the file into .peaks/_runtime/${rdEvidenceDir}/qa/ (the canonical location) so subsequent runs resolve on the canonical path. The legacy \`peaks workspace migrate-change-scope\` helper was removed in v2.19.0; use \`peaks workspace migrate\` to relocate misplaced content.`);
-            }
-        }
-        else {
-            gate.detail = `missing: ${canonicalQaPath}`;
-            violations.push(`QA evidence missing: ${gate.description} (${fileName})`);
-            nextActions.push(`Create .peaks/_runtime/${rdEvidenceDir}/qa/${fileName}`);
-        }
-    }
-    // Check if QA reached verdict-issued
-    if (qaInvoked && !QA_COMPLETE_STATES.has(qaState)) {
-        violations.push(`QA not complete: state is "${qaState}" — must reach "verdict-issued" (functional + performance + security checks done)`);
-        nextActions.push(`Complete QA gates → peaks request transition ${options.rid} --role qa --state verdict-issued`);
-    }
-    // RD invoked without QA — check is moved to AFTER markIfSkipped
-    // (slice 2026-06-13-peaks-workflow-skip) because the gate.status
-    // values are only final after the post-process pass. We track the
-    // decision with a placeholder here and resolve it below.
-    const rdInvokedWithoutQaRaw = rdInvoked && !qaInvoked;
-    // Slice 2026-06-13-peaks-workflow-skip: post-process. For any gate
-    // the user marked as skipped, override the evaluation result. The
-    // boolean `passed` is set to true so the gate counts as satisfied;
-    // the new `status: 'skipped'` field signals the bypass to
-    // downstream consumers. Violations and nextActions pushed for
-    // skipped gates are filtered out (the user explicitly chose to
-    // skip — the missing-evidence message is no longer actionable).
-    for (const gate of [...rdGates, ...qaGates]) {
-        if (skippedGateNames.has(gate.name)) {
-            markIfSkipped(gate);
-        }
-    }
-    if (skippedGateNames.size > 0) {
-        const skippedViolations = new Set();
-        for (const gate of [...rdGates, ...qaGates]) {
-            if (gate.status === 'skipped') {
-                skippedViolations.add(`RD evidence missing: ${gate.description}`);
-                skippedViolations.add(`QA evidence missing: ${gate.description}`);
-            }
-        }
-        for (let i = violations.length - 1; i >= 0; i -= 1) {
-            const v = violations[i];
-            for (const sv of skippedViolations) {
-                if (v.startsWith(sv)) {
-                    violations.splice(i, 1);
-                    break;
-                }
-            }
-        }
-    }
-    // Resolve the "RD invoked without QA" check now that markIfSkipped
-    // has run. The check is suppressed when every QA gate is skipped
-    // (slice 2026-06-13-peaks-workflow-skip: the user explicitly chose
-    // to skip the QA phase).
-    if (rdInvokedWithoutQaRaw) {
-        const allQaSkipped = qaGates.every((g) => g.status === 'skipped');
-        if (!allQaSkipped) {
-            violations.push('CRITICAL: peaks-rd was invoked but peaks-qa was NOT — QA functional/performance/security testing is mandatory after all RD work');
-            nextActions.push('MUST invoke Skill(skill="peaks-qa") before declaring workflow complete');
-        }
-    }
-    // Slice 002 (v2.15.0) AC-3 — Gate H "feedback-promotion". Scans
-    // `.peaks/memory/*.md` for `metadata.type === 'feedback'` entries
-    // without a promotion marker (HTML comment or `.promotion.json`
-    // sidecar). When any unpromoted feedback is found, the gate fails
-    // and the pipeline does not complete until the user promotes via
-    // `peaks feedback promote <memory-file> --layer <A|B|C>`.
-    //
-    // The scan is intentionally non-throwing — a missing or unreadable
-    // memory dir is treated as "no feedback found, gate passes" so
-    // empty projects / pre-feedback-epoch projects don't false-positive.
-    const feedbackGates = [
-        {
-            name: 'feedback-promotion',
-            description: 'Every feedback memory is promoted to at least one enforcement layer (sop / hooks / hard-floor)',
-            passed: false,
-            detail: ''
-        }
-    ];
-    let unpromotedFeedbackCount = 0;
-    try {
-        const unpromoted = listUnpromotedFeedback({ projectRoot: options.projectRoot });
-        unpromotedFeedbackCount = unpromoted.length;
-        if (unpromoted.length === 0) {
-            feedbackGates[0].passed = true;
-            feedbackGates[0].detail = `0 unpromoted feedback memories in .peaks/memory/`;
-        }
-        else {
-            feedbackGates[0].detail = `${unpromoted.length} unpromoted feedback memor${unpromoted.length === 1 ? 'y' : 'ies'}: ${unpromoted.map((u) => u.name).join(', ')}`;
-            violations.push(`Gate H feedback-promotion FAILED: ${unpromoted.length} feedback memor${unpromoted.length === 1 ? 'y is' : 'ies are'} not yet promoted to an enforcement layer (${unpromoted.map((u) => u.name).join(', ')}). Run \`peaks feedback promote <memory-file> --layer <A|B|C>\` for each. See sops/feedback-promotion-sop.md.`);
-            nextActions.push(`Run \`peaks feedback promote <memory-file> --layer <A|B|C>\` for each unpromoted feedback memory to satisfy Gate H.`);
-        }
-    }
-    catch {
-        // listUnpromotedFeedback swallows IO errors internally; the
-        // outer catch is a belt-and-braces guard for unexpected
-        // failures (e.g. permission denied). Treat as "no feedback" —
-        // fail-open on gate infrastructure rather than blocking ship.
-        feedbackGates[0].passed = true;
-        feedbackGates[0].detail = 'feedback-promotion scan skipped (memory dir unreadable; treating as no feedback)';
-    }
-    const allFeedbackGatesPassed = feedbackGates.every((g) => g.passed);
-    const allRdGatesPassed = rdGates.every((g) => g.passed);
-    const allQaGatesPassed = qaGates.every((g) => g.passed);
-    const complete = rdInvoked && qaInvoked && allRdGatesPassed && allQaGatesPassed && allFeedbackGatesPassed
-        && RD_QA_HANDOFF_STATES.has(rdState) && QA_COMPLETE_STATES.has(qaState);
-    // Slice 025 — derive the `acceptedForm` and `gateC` verdict. The form is
-    // 'suffixed' if both the security + perf gates passed via the new
-    // per-rid path; 'legacy' if either was consumed via the legacy fallback;
-    // 'none' if neither passed.
-    const secGate = qaGates.find((g) => g.name === 'security-findings');
-    const perfGate = qaGates.find((g) => g.name === 'performance-findings');
-    const secForm = secGate?.detail?.includes(`-${options.rid}.md`) ? 'suffixed' : 'legacy';
-    const perfForm = perfGate?.detail?.includes(`-${options.rid}.md`) ? 'suffixed' : 'legacy';
-    const acceptedForm = !secGate?.passed && !perfGate?.passed
-        ? 'none'
-        : (secForm === 'suffixed' && perfForm === 'suffixed')
-            ? 'suffixed'
-            : (secForm === 'legacy' || perfForm === 'legacy')
-                ? 'legacy'
-                : 'suffixed';
-    const gateC = allQaGatesPassed ? 'pass' : 'fail';
-    const gateH = allFeedbackGatesPassed ? 'pass' : 'fail';
-    // Slice 2026-06-28-solo-mode-bypass-fix (defect #3): `true` when
-    // every gate resolved on the canonical path; `false` when at least
-    // one fell back to a legacy form. QA / TXT surface the value so the
-    // user knows the legacy content must be moved into the canonical
-    // location (the legacy `peaks workspace migrate-change-scope` helper
-    // was removed in v2.19.0; use `peaks workspace migrate` to relocate
-    // misplaced content).
-    //
-    // Implementation note: use the in-loop flag `allResolvedPathsCanonical`
-    // (correctly updated by both the inline evidence loop AND the
-    // artifact-paths resolver for security/performance findings), not a
-    // string-scan over `gate.detail`. The string-scan approach misses the
-    // artifact-paths legacy branch (it does NOT inject the
-    // `DEPRECATION_LEGACY_PATH_USED` suffix into gate.detail).
-    //
-    // If NO evidence file resolved (e.g. all gates missing), the field is
-    // `true` — there is nothing non-canonical to worry about. If every
-    // resolved gate was skipped via `peaks workflow skip`, the field is
-    // `true` — the user opted out of the canonical/legacy decision.
-    const allGatesSkipped = [...rdGates, ...qaGates].every((g) => g.status === 'skipped');
-    const usedCanonicalPath = !anyEvidenceResolved || allGatesSkipped || allResolvedPathsCanonical;
-    return {
-        rid: options.rid,
-        sessionId: resolvedChangeId,
-        requestType,
-        complete,
-        rdPhase: { invoked: rdInvoked, state: rdState, gates: rdGates },
-        qaPhase: { invoked: qaInvoked, state: qaState, gates: qaGates },
-        feedbackPhase: { gates: feedbackGates },
-        violations,
-        nextActions,
-        acceptedForm,
-        gateC,
-        gateH,
-        usedCanonicalPath
-    };
-}

package/dist/src/services/workflow/workflow-autonomous-service.d.ts

@@ -1,100 +0,0 @@
-import type { CapabilityItemType } from '../recommendations/recommendation-types.js';
-import type { ModelProviderConfig, WorkspaceConfig } from '../config/config-types.js';
-import { type RdPlanResult } from '../rd/rd-service.js';
-import { type SoloMode, type WorkflowMode, type WorkflowRouterPlan } from './workflow-router-service.js';
-export { getResumeRequiredArtifacts, getResumeArtifactsStatus, createResumePlan } from './workflow-autonomous-resume-helpers.js';
-export type CapabilitySurface = 'skill' | 'mcp' | 'plugin' | 'expert';
-export type CapabilityPurpose = 'code-review' | 'security-review' | 'coding-standards' | 'docs-lookup' | 'browser-validation' | 'browser-debug' | 'design-context' | 'design-source' | 'code-search' | 'database-inspection' | 'browser-agent' | 'worker-guidance' | 'memory' | 'context-management' | 'ui-components' | 'spec-workflow' | 'repo-intelligence' | 'openspec' | 'workflow-methodology' | 'workflow-reference' | 'workflow-guidance' | 'product-guidance' | 'design-reference' | 'ui-reference' | 'engineering-guidance' | 'typescript-guidance' | 'quality-guidance' | 'skill-pack' | 'external-skill' | 'design-critique' | 'design-guidance' | 'cloud-skill-pack';
-export type CapabilityActivation = 'available' | 'needs-install' | 'needs-credentials' | 'not-active';
-export type CapabilityTrustLevel = 'local' | 'user-curated' | 'third-party';
-export type CapabilityCandidate = {
-    readonly id: string;
-    readonly source: string;
-    readonly purpose: CapabilityPurpose;
-    readonly surface: CapabilitySurface;
-    readonly kind: CapabilitySurface;
-    readonly sourceType: CapabilityItemType;
-    readonly trustLevel: CapabilityTrustLevel;
-    readonly activation: CapabilityActivation;
-    readonly risk: readonly string[];
-};
-export type AutonomousWorkflowRequest = {
-    readonly mode: WorkflowMode;
-    readonly soloMode?: SoloMode;
-    readonly sessionId: string;
-    readonly goal: string;
-    readonly maxWorkers?: number;
-    readonly dryRun: true;
-    readonly artifactWorkspacePath?: string;
-    readonly workspace?: WorkspaceConfig;
-    readonly config?: {
-        readonly economyMode?: boolean;
-        readonly swarmMode?: boolean;
-        readonly providers?: ModelProviderConfig;
-    };
-};
-export type AutonomousGoalPackage = {
-    readonly sessionId: string;
-    readonly goal: string;
-    readonly nonGoals: readonly string[];
-    readonly preservedBehavior: readonly string[];
-    readonly acceptanceCriteria: readonly string[];
-    readonly doneCondition: string;
-    readonly resumeCondition: string;
-    readonly riskNotes: readonly string[];
-};
-export type AutonomousCapabilityPlan = {
-    readonly sources: readonly string[];
-    readonly policy: readonly string[];
-    readonly candidates: readonly CapabilityCandidate[];
-    readonly surfaces: readonly CapabilitySurface[];
-    readonly surfaceSummary: Record<CapabilitySurface, number>;
-};
-export type AutonomousResumePlan = {
-    readonly status: 'preview' | 'ready';
-    readonly checkpoints: readonly string[];
-    readonly requiredArtifacts: readonly string[];
-    readonly resumeInstructions: string;
-};
-export type AutonomousGoalCommand = {
-    readonly command: string;
-    readonly durable: false;
-    readonly reason: string;
-};
-export type AutonomousStoragePlan = {
-    readonly scope: 'user-local';
-    readonly artifactWorkspacePath: string | null;
-    readonly memoryBackupPath: string | null;
-};
-export type AutonomousMvpPackage = {
-    readonly mode: WorkflowMode;
-    readonly soloMode: SoloMode | undefined;
-    readonly executionMode: 'preview';
-    readonly dryRun: true;
-    readonly routePolicy: WorkflowRouterPlan['routePolicy'];
-    readonly rdWaveNames: readonly string[];
-    readonly capabilitySurfaces: readonly CapabilitySurface[];
-    readonly capabilityCountBySurface: Record<CapabilitySurface, number>;
-    readonly ready: boolean;
-};
-export type AutonomousWorkflowPlan = {
-    readonly available: boolean;
-    readonly behavior: 'preview' | 'ready';
-    readonly sessionId: string;
-    readonly goal: string;
-    readonly mode: WorkflowMode;
-    readonly dryRun: true;
-    readonly goalPackage: AutonomousGoalPackage;
-    readonly goalCommand: AutonomousGoalCommand;
-    readonly capabilityPlan: AutonomousCapabilityPlan;
-    readonly storagePlan: AutonomousStoragePlan;
-    readonly routePlan: WorkflowRouterPlan;
-    readonly modelAssignments: WorkflowRouterPlan['modelAssignments'];
-    readonly rdPlan: RdPlanResult;
-    readonly resumePlan: AutonomousResumePlan;
-    readonly mvpPackage: AutonomousMvpPackage;
-    readonly constraints: readonly string[];
-    readonly blockedReasons: readonly string[];
-    readonly nextActions: readonly string[];
-};
-export declare function createAutonomousWorkflowPlan(request: AutonomousWorkflowRequest): AutonomousWorkflowPlan;