wogiflow 2.26.2 → 2.29.0

package/lib/workspace-task-injector.js

@@ -0,0 +1,223 @@
+'use strict';
+
+/**
+ * Wogi Workspace — Manager-side Task Injector (wf-2f49b292 / G5)
+ *
+ * Writes a task record into a worker's `.workflow/state/ready.json` from the
+ * manager, so the manager can dispatch `/wogi-start <taskId>` for a brand-new
+ * task that the worker doesn't yet know about.
+ *
+ * Before G5: manager could only dispatch tasks that already existed in the
+ * worker's ready.json (worker had to create them first). G5 closes that gap.
+ *
+ * Atomicity: write-to-temp + rename — atomic at the filesystem layer.
+ * Concurrent injects from two manager turns serialize via rename semantics;
+ * the last writer wins only on the rename, so we re-read before every write
+ * to avoid losing a concurrent append.
+ */
+
+const fs = require('node:fs');
+const path = require('node:path');
+const crypto = require('node:crypto');
+
+const VALID_REPO_NAME = /^[a-zA-Z0-9_-]{1,64}$/;
+const VALID_TASK_ID = /^wf-[0-9a-f]{8}$/i;
+const REQUIRED_FIELDS = ['id', 'title', 'type'];
+const MAX_RETRIES = 5;
+
+/**
+ * Resolve the filesystem path to a worker's ready.json.
+ * Reads wogi-workspace.json manifest to find the worker's repo path.
+ *
+ * @param {string} workspaceRoot
+ * @param {string} repoName
+ * @returns {string} absolute path to ready.json (may not yet exist)
+ */
+function getWorkerReadyPath(workspaceRoot, repoName) {
+  if (!workspaceRoot || typeof workspaceRoot !== 'string') {
+    throw new Error('workspaceRoot must be a non-empty string');
+  }
+  if (!VALID_REPO_NAME.test(repoName)) {
+    throw new Error(`Invalid repoName: "${repoName}" — must match ${VALID_REPO_NAME}`);
+  }
+
+  const configPath = path.join(workspaceRoot, 'wogi-workspace.json');
+  let manifest;
+  try {
+    manifest = JSON.parse(fs.readFileSync(configPath, 'utf-8'));
+  } catch (err) {
+    throw new Error(`Cannot read workspace manifest at ${configPath}: ${err.message}`);
+  }
+
+  const member = manifest.members?.[repoName];
+  if (!member) {
+    throw new Error(`Unknown repo "${repoName}" in workspace manifest`);
+  }
+
+  const memberPath = member.path || `./${repoName}`;
+  const repoPath = path.resolve(workspaceRoot, memberPath);
+
+  const resolvedRoot = path.resolve(workspaceRoot);
+  if (!repoPath.startsWith(resolvedRoot + path.sep) && repoPath !== resolvedRoot) {
+    throw new Error(`Worker repo path escapes workspace root: ${repoPath}`);
+  }
+
+  return path.join(repoPath, '.workflow', 'state', 'ready.json');
+}
+
+/**
+ * Validate a task record has the shape needed for ready.json entries.
+ *
+ * @param {Object} taskRecord
+ * @throws {Error} if invalid
+ */
+function validateTaskRecord(taskRecord) {
+  if (!taskRecord || typeof taskRecord !== 'object' || Array.isArray(taskRecord)) {
+    throw new Error('taskRecord must be a plain object');
+  }
+  for (const field of REQUIRED_FIELDS) {
+    if (!taskRecord[field] || typeof taskRecord[field] !== 'string') {
+      throw new Error(`taskRecord.${field} is required and must be a string`);
+    }
+  }
+  if (!VALID_TASK_ID.test(taskRecord.id)) {
+    throw new Error(`Invalid task ID "${taskRecord.id}" — expected wf-XXXXXXXX (8 hex)`);
+  }
+  if (taskRecord.title.length > 500) {
+    throw new Error('taskRecord.title exceeds 500 chars');
+  }
+}
+
+function atomicWriteJson(filePath, data) {
+  const dir = path.dirname(filePath);
+  fs.mkdirSync(dir, { recursive: true });
+  const tmp = path.join(dir, `.${path.basename(filePath)}.${crypto.randomBytes(6).toString('hex')}.tmp`);
+  fs.writeFileSync(tmp, JSON.stringify(data, null, 2));
+  fs.renameSync(tmp, filePath);
+}
+
+function readReadyJson(readyPath) {
+  if (!fs.existsSync(readyPath)) {
+    return {
+      lastUpdated: new Date().toISOString(),
+      inProgress: [],
+      ready: [],
+      blocked: [],
+      recentlyCompleted: []
+    };
+  }
+  const raw = fs.readFileSync(readyPath, 'utf-8');
+  const parsed = JSON.parse(raw);
+  if (!parsed || typeof parsed !== 'object') {
+    throw new Error(`ready.json is not a valid object at ${readyPath}`);
+  }
+  parsed.ready = Array.isArray(parsed.ready) ? parsed.ready : [];
+  parsed.inProgress = Array.isArray(parsed.inProgress) ? parsed.inProgress : [];
+  parsed.blocked = Array.isArray(parsed.blocked) ? parsed.blocked : [];
+  parsed.recentlyCompleted = Array.isArray(parsed.recentlyCompleted) ? parsed.recentlyCompleted : [];
+  return parsed;
+}
+
+function taskExistsAnywhere(data, taskId) {
+  const lists = ['inProgress', 'ready', 'blocked', 'recentlyCompleted'];
+  for (const k of lists) {
+    if (data[k].some(t => t && t.id === taskId)) return k;
+  }
+  return null;
+}
+
+/**
+ * Inject a task record into a worker's ready.json `ready[]` array.
+ *
+ * Idempotent: if a task with the same id already exists anywhere in the file,
+ * returns `{ ok: true, alreadyPresent: <list-name> }` without modifying the file.
+ *
+ * @param {string} workspaceRoot
+ * @param {string} repoName
+ * @param {Object} taskRecord — must have id (wf-XXXXXXXX), title, type
+ * @returns {{ ok: boolean, path?: string, taskId?: string, alreadyPresent?: string, message?: string }}
+ */
+function injectTask(workspaceRoot, repoName, taskRecord) {
+  try {
+    validateTaskRecord(taskRecord);
+  } catch (err) {
+    return { ok: false, message: err.message };
+  }
+
+  let readyPath;
+  try {
+    readyPath = getWorkerReadyPath(workspaceRoot, repoName);
+  } catch (err) {
+    return { ok: false, message: err.message };
+  }
+
+  for (let attempt = 0; attempt < MAX_RETRIES; attempt++) {
+    let data;
+    try {
+      data = readReadyJson(readyPath);
+    } catch (err) {
+      return { ok: false, message: `Read failed: ${err.message}` };
+    }
+
+    const existingList = taskExistsAnywhere(data, taskRecord.id);
+    if (existingList) {
+      return { ok: true, path: readyPath, taskId: taskRecord.id, alreadyPresent: existingList };
+    }
+
+    const enriched = {
+      ...taskRecord,
+      status: taskRecord.status || 'ready',
+      created: taskRecord.created || new Date().toISOString(),
+      injectedBy: process.env.WOGI_REPO_NAME || 'manager',
+      injectedAt: new Date().toISOString()
+    };
+    data.ready.push(enriched);
+    data.lastUpdated = new Date().toISOString();
+
+    try {
+      atomicWriteJson(readyPath, data);
+      return { ok: true, path: readyPath, taskId: taskRecord.id };
+    } catch (err) {
+      if (attempt === MAX_RETRIES - 1) {
+        return { ok: false, message: `Write failed after ${MAX_RETRIES} attempts: ${err.message}` };
+      }
+    }
+  }
+
+  return { ok: false, message: 'Unreachable retry exhaustion' };
+}
+
+/**
+ * Inject a task into the worker's ready.json, then dispatch /wogi-start.
+ *
+ * If injection reports `alreadyPresent`, still proceeds to dispatch — the
+ * task exists, so dispatch is the right next step.
+ *
+ * @param {string} workspaceRoot
+ * @param {string} repoName
+ * @param {Object} taskRecord
+ * @param {Object} [opts] — forwarded to dispatchToChannel
+ * @returns {Promise<{ inject: Object, dispatch: Object, ok: boolean }>}
+ */
+async function injectAndDispatch(workspaceRoot, repoName, taskRecord, opts = {}) {
+  const injectResult = injectTask(workspaceRoot, repoName, taskRecord);
+  if (!injectResult.ok) {
+    return { ok: false, inject: injectResult, dispatch: null };
+  }
+
+  const { dispatchToChannel } = require('./workspace-routing');
+  const dispatchResult = await dispatchToChannel(workspaceRoot, repoName, taskRecord.id, opts);
+
+  return {
+    ok: injectResult.ok && dispatchResult.ok,
+    inject: injectResult,
+    dispatch: dispatchResult
+  };
+}
+
+module.exports = {
+  injectTask,
+  injectAndDispatch,
+  getWorkerReadyPath,
+  validateTaskRecord
+};

package/lib/workspace.js

@@ -1216,6 +1216,29 @@ You are a workspace worker. There is NO human watching your terminal. You MUST o
 
 The \`TaskCompleted\` hook will inject an auto-pickup directive when channel dispatches are queued (v2.20.0+). The \`Stop\` hook will BLOCK end-of-turn if you try to stop while dispatches are queued and no task is in progress. These enforcements exist because the silent-stall pattern was incident-worthy (2026-04-16).
 
+### Tool-First Turn Contract (v2.27.0+)
+
+**Every worker turn after a UserPromptSubmit MUST contain at least one tool call. In strict mode (default), the FIRST content block must be a tool call, not text.**
+
+Rule name: \`worker-tool-first-turn\`. The \`Stop\` hook blocks end-of-turn when this contract is violated:
+
+- **silent-halt**: zero tool calls across the entire turn (pure-text response)
+- **text-before-tool-call**: first content block is text, not tool_use (strict mode only)
+
+Why: pure-text worker responses are invisible to the user — the user only sees the manager terminal. Worker text disappears into the transcript with no downstream consumer. It also disqualifies the three-state end-of-turn contract (ACTION | ESCALATION | IDLE) above.
+
+**Allowed turn shapes**:
+- \`tool_use → tool_use → end\`
+- \`tool_use → text → tool_use → end\` (narrate after acting)
+- \`tool_use (## QUESTION: dispatch) → end\` (escalation)
+- \`tool_use (## Results: dispatch) → end\` (reply)
+
+**Blocked turn shapes**:
+- \`text → end\` (silent-halt)
+- \`text → tool_use → end\` (text-before-tool-call — strict mode)
+
+Config: \`workspace.toolFirstTurnGate.{enabled, strict}\` in \`.workflow/config.json\`. Default \`enabled: true, strict: true\`.
+
 ### CRITICAL: Stop, Don't Degrade
 
 **If you cannot verify your work to the required evidence tier, you may NOT mark the task as complete.** Report it as BLOCKED with the specific verification gap.

package/lib/worktree-review.js

@@ -0,0 +1,315 @@
+'use strict';
+
+/**
+ * Wogi Flow — Parallel-Worktree Auto Review (wf-8d635d0e / E1)
+ *
+ * On `coding → validating` phase transition, a background "review worker"
+ * scans the task's changes in an isolated git worktree and writes findings
+ * to `.workflow/state/auto-review-findings.json`. The Completion Truth Gate
+ * reads the file and downgrades completion claims when high-severity findings
+ * are present.
+ *
+ * Architecture:
+ *   - `startReview({ taskId })` — detached background spawn (fire-and-forget)
+ *   - `runReview({ taskId, reviewer, ... })` — core scan loop (used in-process
+ *     by the detached child; injectable `reviewer` for tests)
+ *   - `writeFindings / readFindings / awaitFindings` — JSON file I/O helpers.
+ *
+ * The "workspace worker" per the spec is a detached Node child process — not
+ * an MCP-channel-hosted Claude session. Per-task model selection is a Claude
+ * Code feature we don't yet own (AC6 — documented known limitation); the
+ * review therefore defaults to a heuristic static scan of the worktree diff
+ * so it's useful today without a running Claude Code child.
+ */
+
+const path = require('node:path');
+const { spawn, execFileSync } = require('node:child_process');
+
+const { PATHS } = require('../scripts/flow-paths');
+const { readJson, writeJson } = require('../scripts/flow-io');
+
+const FINDINGS_FILE = path.join(PATHS.state, 'auto-review-findings.json');
+const DEFAULT_TIMEOUT_MS = 90000;
+const POLL_INTERVAL_MS = 250;
+
+// ============================================================
+// File I/O — { schemaVersion, records: [...] } object wrapping the records
+// array. Object-at-root is required by safeJsonParse (prototype-pollution
+// protection). Semantics per AC3 remain "array; last-write-wins per task":
+// `records` is the array, one entry per taskId after last-write dedupe.
+// ============================================================
+
+const SCHEMA_VERSION = 1;
+
+function readAll() {
+  const data = readJson(FINDINGS_FILE, { schemaVersion: SCHEMA_VERSION, records: [] });
+  if (data && Array.isArray(data.records)) return data.records;
+  return [];
+}
+
+function readFindings(taskId) {
+  if (!taskId) return null;
+  const all = readAll();
+  // Latest record wins (iterate reverse).
+  for (let i = all.length - 1; i >= 0; i--) {
+    if (all[i] && all[i].taskId === taskId) return all[i];
+  }
+  return null;
+}
+
+/**
+ * Upsert a record for `taskId`. Last-write-wins per AC3: the existing record
+ * (if any) is replaced by `record`.
+ */
+function writeFindings(record) {
+  if (!record || !record.taskId) {
+    throw new Error('writeFindings requires { taskId, ... }');
+  }
+  const all = readAll();
+  const filtered = all.filter((r) => r && r.taskId !== record.taskId);
+  filtered.push(record);
+  writeJson(FINDINGS_FILE, { schemaVersion: SCHEMA_VERSION, records: filtered });
+  return record;
+}
+
+/**
+ * Poll for a terminal record (status !== 'in-progress') for `taskId` within
+ * `timeoutMs`. Returns either the completed record or a synthetic
+ * `unverified-review-timeout` record. Does NOT modify the findings file on
+ * timeout — that's the caller's call.
+ */
+async function awaitFindings(taskId, timeoutMs = DEFAULT_TIMEOUT_MS) {
+  const deadline = Date.now() + Math.max(0, timeoutMs);
+  for (;;) {
+    const rec = readFindings(taskId);
+    if (rec && rec.status && rec.status !== 'in-progress') return rec;
+    if (Date.now() >= deadline) {
+      return {
+        taskId,
+        status: 'unverified-review-timeout',
+        startedAt: rec?.startedAt || null,
+        completedAt: null,
+        findings: [],
+        timeoutMs,
+      };
+    }
+    await sleep(POLL_INTERVAL_MS);
+  }
+}
+
+function sleep(ms) {
+  return new Promise((r) => setTimeout(r, ms));
+}
+
+// ============================================================
+// Heuristic default reviewer (no Claude dependency)
+// ============================================================
+
+const FINDING_PATTERNS = [
+  { re: /\bTODO\b/, severity: 'low', claim: 'TODO left in changed code' },
+  { re: /\bFIXME\b/, severity: 'medium', claim: 'FIXME left in changed code' },
+  { re: /\bXXX\b/, severity: 'medium', claim: 'XXX marker left in changed code' },
+  { re: /^<<<<<<<|^=======$|^>>>>>>>/, severity: 'high', claim: 'Unresolved merge conflict marker' },
+  { re: /console\.log\(/, severity: 'low', claim: 'console.log left in code' },
+  { re: /debugger;/, severity: 'high', claim: 'debugger statement left in code' },
+];
+
+/**
+ * Default reviewer — runs in the worktree and reports findings on changed
+ * lines only (git diff scope). Pure static scan; no AI. Returns findings[].
+ *
+ * This is intentionally cheap so the feature is useful without per-task
+ * Claude model selection (AC6). When that Claude Code feature lands, a
+ * richer reviewer can be plugged in via the `reviewer` option of runReview.
+ */
+function heuristicReviewer({ worktreePath, baseBranch }) {
+  const findings = [];
+  let diff = '';
+  try {
+    diff = execFileSync('git', ['diff', `${baseBranch}...HEAD`, '--unified=0'], {
+      cwd: worktreePath,
+      encoding: 'utf-8',
+      maxBuffer: 16 * 1024 * 1024,
+    });
+  } catch (_err) {
+    return findings;
+  }
+  if (!diff) return findings;
+
+  let currentFile = null;
+  let currentLine = 0;
+  for (const raw of diff.split('\n')) {
+    if (raw.startsWith('+++ b/')) {
+      currentFile = raw.slice(6);
+      currentLine = 0;
+      continue;
+    }
+    if (raw.startsWith('@@')) {
+      // @@ -a,b +c,d @@
+      const m = raw.match(/\+(\d+)/);
+      currentLine = m ? parseInt(m[1], 10) : 0;
+      continue;
+    }
+    if (!raw.startsWith('+') || raw.startsWith('+++')) continue;
+    const added = raw.slice(1);
+    for (const pat of FINDING_PATTERNS) {
+      if (pat.re.test(added)) {
+        findings.push({
+          severity: pat.severity,
+          file: currentFile,
+          line: currentLine,
+          claim: pat.claim,
+          evidence: added.trim().slice(0, 200),
+        });
+      }
+    }
+    currentLine += 1;
+  }
+  return findings;
+}
+
+// ============================================================
+// Run (in-process) — the detached child calls this
+// ============================================================
+
+/**
+ * Run a review for `taskId` against `worktreePath`. Updates the findings file
+ * with `in-progress` → `complete|error` status transitions. Returns the final
+ * record.
+ *
+ * @param {Object} opts
+ * @param {string} opts.taskId
+ * @param {string} opts.worktreePath
+ * @param {string} [opts.baseBranch='HEAD~1']
+ * @param {Function} [opts.reviewer] — (ctx) => findings[] | Promise<findings[]>.
+ *   Injectable for tests. Default = heuristicReviewer.
+ */
+async function runReview({ taskId, worktreePath, baseBranch = 'HEAD~1', reviewer = heuristicReviewer }) {
+  if (!taskId) throw new Error('runReview requires taskId');
+  const startedAt = new Date().toISOString();
+  writeFindings({
+    taskId,
+    status: 'in-progress',
+    startedAt,
+    completedAt: null,
+    findings: [],
+  });
+
+  try {
+    const findings = await Promise.resolve(reviewer({ taskId, worktreePath, baseBranch }));
+    const record = {
+      taskId,
+      status: 'complete',
+      startedAt,
+      completedAt: new Date().toISOString(),
+      findings: Array.isArray(findings) ? findings : [],
+    };
+    writeFindings(record);
+    return record;
+  } catch (err) {
+    const record = {
+      taskId,
+      status: 'error',
+      startedAt,
+      completedAt: new Date().toISOString(),
+      findings: [],
+      error: String(err.message || err),
+    };
+    writeFindings(record);
+    return record;
+  }
+}
+
+// ============================================================
+// startReview — the public entry (spawn detached child)
+// ============================================================
+
+/**
+ * Fire-and-forget background review. Creates a worktree, spawns a detached
+ * child that runs `runReview`, and returns a handle the caller can poll via
+ * `awaitFindings(taskId, timeoutMs)`. The child is unref()d so it never
+ * keeps the parent process alive.
+ *
+ * Non-blocking per AC5: the parent writes an `in-progress` marker and
+ * returns immediately. If the child never completes within `timeoutMs`,
+ * `awaitFindings` returns an `unverified-review-timeout` record.
+ *
+ * @param {Object} opts
+ * @param {string} opts.taskId
+ * @param {string} [opts.repoRoot] — default = process.cwd()
+ * @param {string} [opts.childScript] — override the worker entry (tests)
+ * @returns {{ taskId, pid, worktreePath }}
+ */
+function startReview({ taskId, repoRoot, childScript }) {
+  if (!taskId) throw new Error('startReview requires taskId');
+  const root = repoRoot || process.cwd();
+
+  // Write the in-progress marker synchronously so consumers see it immediately.
+  writeFindings({
+    taskId,
+    status: 'in-progress',
+    startedAt: new Date().toISOString(),
+    completedAt: null,
+    findings: [],
+  });
+
+  // Worktree creation happens inside the child process (the worker script
+  // is responsible for createWorktree / discardWorktree — keeping this
+  // parent call synchronous and fire-and-forget).
+  const worktreePath = null;
+
+  const entry = childScript || path.join(__dirname, '..', 'scripts', 'flow-auto-review-worker.js');
+  const child = spawn(process.execPath, [entry, '--task', taskId, '--repoRoot', root], {
+    cwd: root,
+    detached: true,
+    stdio: 'ignore',
+    env: { ...process.env, WOGI_AUTO_REVIEW_CHILD: '1' },
+  });
+  child.unref();
+
+  return { taskId, pid: child.pid, worktreePath };
+}
+
+// ============================================================
+// Findings → audit incorporation (consumed by completion-truth-gate)
+// ============================================================
+
+/**
+ * Summarize findings for a task into a shape the truth gate can fold into its
+ * existing `audit` object. High-severity findings trigger a soft-warn or block
+ * (depending on configuration) without the gate having to re-implement any
+ * downgrade logic — it just treats this like another "insufficient" signal.
+ */
+function summarizeFindingsForAudit(taskId) {
+  const rec = readFindings(taskId);
+  if (!rec) return { present: false };
+
+  const findings = Array.isArray(rec.findings) ? rec.findings : [];
+  const counts = { low: 0, medium: 0, high: 0 };
+  for (const f of findings) {
+    const sev = (f && f.severity) || 'low';
+    if (counts[sev] !== undefined) counts[sev] += 1;
+  }
+
+  return {
+    present: true,
+    status: rec.status || 'unknown',
+    timedOut: rec.status === 'unverified-review-timeout',
+    counts,
+    highSeverityCount: counts.high,
+    topHighSeverity: findings.filter((f) => f && f.severity === 'high').slice(0, 5),
+  };
+}
+
+module.exports = {
+  FINDINGS_FILE,
+  DEFAULT_TIMEOUT_MS,
+  readFindings,
+  writeFindings,
+  awaitFindings,
+  runReview,
+  startReview,
+  heuristicReviewer,
+  summarizeFindingsForAudit,
+  _internal: { readAll },
+};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wogiflow",
-  "version": "2.26.2",
+  "version": "2.29.0",
   "description": "AI-powered development workflow management system with multi-model support",
   "main": "lib/index.js",
   "bin": {
@@ -10,7 +10,7 @@
   },
   "scripts": {
     "flow": "./scripts/flow",
-    "test": "NODE_ENV=test node --test tests/auto-compact-prompt.test.js tests/flow-paths.test.js tests/flow-io.test.js tests/flow-config-loader.test.js tests/flow-damage-control.test.js tests/flow-output.test.js tests/flow-constants.test.js tests/flow-session-state.test.js tests/flow-hooks-integration.test.js tests/flow-utils.test.js tests/flow-security.test.js tests/flow-memory-db.test.js tests/flow-durable-session.test.js tests/flow-skill-matcher.test.js tests/flow-bridge.test.js tests/flow-proactive-compact.test.js tests/flow-cascade-completion.test.js tests/flow-capture-gate.test.js tests/flow-correction-detector-hybrid.test.js tests/flow-promote.test.js tests/flow-archive-runs.test.js tests/flow-memory.test.js tests/flow-hooks-pre-tool-helpers.test.js tests/flow-hooks-bugfix-scope-gate.test.js tests/flow-hooks-routing-gate.test.js tests/flow-hooks-phase-read-gate.test.js tests/flow-hooks-commit-log-gate.test.js tests/flow-hooks-deploy-gate.test.js tests/flow-hooks-todowrite-gate.test.js tests/flow-hooks-git-safety-gate.test.js tests/flow-hooks-scope-mutation-gate.test.js tests/flow-hooks-strike-gate.test.js tests/flow-hooks-component-check.test.js tests/flow-hooks-scope-gate.test.js tests/flow-hooks-implementation-gate.test.js tests/flow-hooks-research-gate.test.js tests/flow-hooks-loop-check.test.js tests/flow-hooks-manager-boundary-gate.test.js tests/flow-hooks-phase-gate.test.js tests/flow-hooks-pre-tool-orchestrator.test.js tests/flow-hooks-observation-capture.test.js tests/flow-hooks-task-gate.test.js tests/flow-durable-session-suspension.test.js tests/flow-health-mcp-scopes.test.js tests/flow-lean-config.test.js tests/flow-workspace-autopickup.test.js tests/flow-worker-boundary-gate.test.js tests/flow-worker-question-classifier.test.js tests/flow-completion-truth-gate-contradictions.test.js tests/flow-structure-sensor.test.js tests/flow-workspace-dispatch-tracking.test.js tests/flow-story-gates.test.js tests/flow-workspace-restart-handoff.test.js tests/flow-wogi-claude-wrapper.test.js tests/flow-wave1-integrations.test.js tests/flow-wave2-integrations.test.js tests/flow-wave3-integrations.test.js tests/flow-commit-claims-gate.test.js && NODE_ENV=test node tests/run-quality-gates.test.js",
+    "test": "NODE_ENV=test node --test tests/auto-compact-prompt.test.js tests/flow-paths.test.js tests/flow-io.test.js tests/flow-config-loader.test.js tests/flow-damage-control.test.js tests/flow-output.test.js tests/flow-constants.test.js tests/flow-session-state.test.js tests/flow-hooks-integration.test.js tests/flow-utils.test.js tests/flow-security.test.js tests/flow-memory-db.test.js tests/flow-durable-session.test.js tests/flow-skill-matcher.test.js tests/flow-bridge.test.js tests/flow-proactive-compact.test.js tests/flow-cascade-completion.test.js tests/flow-capture-gate.test.js tests/flow-correction-detector-hybrid.test.js tests/flow-promote.test.js tests/flow-archive-runs.test.js tests/flow-memory.test.js tests/flow-hooks-pre-tool-helpers.test.js tests/flow-hooks-bugfix-scope-gate.test.js tests/flow-hooks-routing-gate.test.js tests/flow-hooks-phase-read-gate.test.js tests/flow-hooks-commit-log-gate.test.js tests/flow-hooks-deploy-gate.test.js tests/flow-hooks-todowrite-gate.test.js tests/flow-hooks-git-safety-gate.test.js tests/flow-hooks-scope-mutation-gate.test.js tests/flow-hooks-strike-gate.test.js tests/flow-hooks-component-check.test.js tests/flow-hooks-scope-gate.test.js tests/flow-hooks-implementation-gate.test.js tests/flow-hooks-research-gate.test.js tests/flow-hooks-loop-check.test.js tests/flow-hooks-manager-boundary-gate.test.js tests/flow-hooks-phase-gate.test.js tests/flow-hooks-pre-tool-orchestrator.test.js tests/flow-hooks-observation-capture.test.js tests/flow-hooks-task-gate.test.js tests/flow-durable-session-suspension.test.js tests/flow-health-mcp-scopes.test.js tests/flow-lean-config.test.js tests/flow-workspace-autopickup.test.js tests/flow-worker-boundary-gate.test.js tests/flow-worker-question-classifier.test.js tests/flow-completion-truth-gate-contradictions.test.js tests/flow-structure-sensor.test.js tests/flow-workspace-dispatch-tracking.test.js tests/workspace-ipc-sqlite.test.js tests/workspace-ipc-multi-worker.test.js tests/flow-story-gates.test.js tests/flow-workspace-restart-handoff.test.js tests/flow-wogi-claude-wrapper.test.js tests/flow-wave1-integrations.test.js tests/flow-wave2-integrations.test.js tests/flow-wave3-integrations.test.js tests/flow-commit-claims-gate.test.js tests/auto-review.test.js tests/gate-telemetry-surface.test.js tests/agents-md-alias.test.js tests/flow-skill-manage.test.js tests/fuzzy-patch.test.js tests/mode-schema.test.js tests/flow-feature-dossier.test.js && NODE_ENV=test node tests/run-quality-gates.test.js",
     "test:syntax": "find scripts/ lib/ -name '*.js' -not -path '*/node_modules/*' -exec node --check {} +",
     "lint": "eslint scripts/ lib/ tests/",
     "lint:ci": "eslint scripts/ lib/ tests/ --max-warnings 0",

package/scripts/base-workflow-step.js

@@ -77,7 +77,7 @@ class BaseWorkflowStep {
    * @param {object} options - Remaining options (stepConfig, mode, taskType, etc.)
    * @returns {Promise<{passed: boolean, message: string, details?: any}>}
    */
-  async execute(_files, options) {
+  async execute(_files, _options) {
     throw new Error(`${this.name}: execute() must be overridden`);
   }
 

package/scripts/flow

@@ -541,6 +541,12 @@ case "${1:-}" in
     memory)
         node "$SCRIPT_DIR/flow-memory.js" "${@:2}"
         ;;
+    feature-dossier|dossier)
+        node "$SCRIPT_DIR/flow-feature-dossier.js" "${@:2}"
+        ;;
+    logic-rules)
+        node "$SCRIPT_DIR/flow-logic-rules.js" "${@:2}"
+        ;;
     skill-create)
         node "$SCRIPT_DIR/flow-skill-create.js" "${@:2}"
         ;;
@@ -670,8 +676,13 @@ case "${1:-}" in
                 node -e "require('$SCRIPT_DIR/../lib/skill-registry').skill(['add', '$3', ...process.argv.slice(2)])" -- "${@:4}"
                 ;;
             remove)
-                # Remove installed skill
-                node -e "require('$SCRIPT_DIR/../lib/skill-registry').skill(['remove', '$3'])"
+                # If --name is present, this is a staged-removal proposal (flow-skill-manage).
+                # Otherwise, legacy registry removal.
+                if [[ " $* " == *" --name "* ]]; then
+                    node "$SCRIPT_DIR/flow-skill-manage.js" remove "${@:3}"
+                else
+                    node -e "require('$SCRIPT_DIR/../lib/skill-registry').skill(['remove', '$3'])"
+                fi
                 ;;
             update)
                 # Update skills
@@ -685,8 +696,12 @@ case "${1:-}" in
                 # List skills from registry
                 node -e "require('$SCRIPT_DIR/../lib/skill-registry').skill(['list'])"
                 ;;
+            propose|patch|promote|reject|archive|pending)
+                # Agent proposal CLI (staged, user-approved at session-end)
+                node "$SCRIPT_DIR/flow-skill-manage.js" "${@:2}"
+                ;;
             *)
-                echo "Usage: flow skill [detect|list|create|add|remove|update|info|registry]"
+                echo "Usage: flow skill [detect|list|create|add|remove|update|info|registry|propose|patch|promote|reject|archive|pending]"
                 echo ""
                 echo "Local Skills:"
                 echo "  detect              Detect frameworks in project"
@@ -696,9 +711,18 @@ case "${1:-}" in
                 echo "Registry Skills:"
                 echo "  registry            List available skills from registry"
                 echo "  add <name>          Install skill from registry"
-                echo "  remove <name>       Remove installed skill"
+                echo "  remove <name>       Remove installed skill (or stage proposal with --name)"
                 echo "  update [name]       Update skill(s)"
                 echo "  info <name>         Show skill details"
+                echo ""
+                echo "Agent Proposals (staged, user-approved at session-end):"
+                echo "  propose --name <n> --content <f>  Stage new skill"
+                echo "  patch   --name <n> --content <f>  Stage edit to existing skill"
+                echo "  remove  --name <n>                Stage removal of existing skill"
+                echo "  promote <name>                    Apply pending proposal"
+                echo "  reject  <name>                    Discard pending proposal"
+                echo "  archive <name>                    Direct archive (no staging)"
+                echo "  pending [--json]                  List pending proposals"
                 ;;
         esac
         ;;