mindforge-cc 8.1.1 → 8.2.0

package/.planning/archive/v8-cleanup/implementation-roadmap/v2.0.0/day-14-prod-v2/DAY14-V2-PRODUCTION-RELEASE.md

@@ -0,0 +1,2255 @@
+# MindForge v2 — Day 14: v2.0.0 Production Release
+# Branch: `feat/mindforge-v2-release`
+# Prerequisite: `feat/mindforge-v2-self-building-skills` merged to `main`
+# Version target: v2.0.0 — "The Autonomous Enterprise"
+# Theme: "Ship It. Every Promise Kept."
+
+---
+
+## BRANCH SETUP
+
+```bash
+git checkout main
+git pull origin main
+
+# Verify Day 13 baseline — CRITICAL: everything must pass before release work begins
+node -e "console.log(require('./package.json').version)"  # Must be 2.0.0-alpha.6
+
+# Run ALL 21 test suites — zero tolerance for failures before release
+SUITES=(install wave-engine audit compaction skills-platform \
+        integrations governance intelligence metrics \
+        distribution ci-mode sdk production migration e2e \
+        autonomous browser model-routing memory dashboard \
+        self-building-skills)
+
+FAIL=0
+for suite in "${SUITES[@]}"; do
+  printf "  %-35s" "${suite}..."
+  node tests/${suite}.test.js 2>&1 | tail -1 | grep -q "passed" && echo "✅" || { echo "❌ BLOCKED"; ((FAIL++)); }
+done
+
+[ "$FAIL" -gt 0 ] && echo "
+❌ $FAIL test suite(s) failing — FIX BEFORE STARTING DAY 14.
+   v2.0.0 release cannot proceed with failing tests." && exit 1
+
+echo "
+✅ All 21 test suites pass. Day 14 release work may begin."
+
+git checkout -b feat/mindforge-v2-release
+```
+
+---
+
+## DAY 14 SCOPE
+
+Day 14 is **The Release**. Every feature built in Days 8-13 ships as a cohesive
+v2.0.0 package. The scope is different from previous days: instead of building
+new features, Day 14 integrates, validates, hardens, and ships everything built so far.
+
+**The v2.0.0 promise:**
+> "The only agentic framework that is simultaneously enterprise-grade
+> (governance, compliance, integrations) AND autonomously capable
+> (walk-away mode, browser automation, cross-model intelligence,
+> persistent memory, real-time team dashboard, self-building skills)."
+
+**Day 14 deliverables:**
+
+| Deliverable | Description |
+|---|---|
+| Multi-runtime expansion | 6 runtimes: Claude Code, Antigravity, Cursor, OpenCode, Gemini CLI, GitHub Copilot |
+| `/mindforge:new-runtime` command | Scaffold support for any new runtime |
+| v2.0.0 migration script | 1.0.0→2.0.0 HANDOFF.json + AUDIT.jsonl backfill |
+| 65-point production checklist | 50 original + 15 new v2 items (Section F) |
+| Complete v2.0.0 CHANGELOG | Full "The Autonomous Enterprise" entry |
+| Full test battery × 3 runs | All 22 test suites, zero failures, three consecutive passes |
+| `tests/release.test.js` | 22nd and final test suite |
+| v2.0.0 state file | MINDFORGE-STATE-V2.md for post-release continuation |
+| npm pack simulation | Verify tarball integrity before publish |
+| GitHub release notes | Complete release notes document |
+
+**New commands today: 48 total (47 + new-runtime)**
+**Final ADR count: 41 (ADR-001 through ADR-041)**
+
+---
+
+# ═══════════════════════════════════════════════════════════════════════
+# PART 1 — IMPLEMENTATION PROMPT
+# ═══════════════════════════════════════════════════════════════════════
+
+---
+
+## TASK 1 — Multi-Runtime Expansion: Update installer for 6 runtimes
+
+### Update `bin/installer-core.js` RUNTIMES config
+
+The existing installer supports Claude Code and Antigravity. Day 14 adds 4 more:
+Cursor, OpenCode, Gemini CLI, and GitHub Copilot.
+
+```javascript
+/**
+ * MindForge v2.0.0 — RUNTIMES configuration
+ * 6 runtimes supported: Claude Code, Antigravity, Cursor,
+ * OpenCode, Gemini CLI, GitHub Copilot.
+ *
+ * Each runtime definition:
+ *   globalDir      — Machine-level directory for global installations
+ *   localDir       — Project-level directory (relative to project root)
+ *   commandsSubdir — Subdirectory within localDir for command files
+ *   entryFile      — Main agent instructions file name
+ *   commandFormat  — How this runtime reads commands ('slash' | 'file' | 'include')
+ *   supportsSlash  — Whether /mindforge:cmd syntax is used directly
+ */
+
+// Replace the existing RUNTIMES const in bin/installer-core.js with:
+
+const RUNTIMES = {
+
+  // ── Claude Code ─────────────────────────────────────────────────────────────
+  claude: {
+    id:             'claude',
+    displayName:    'Claude Code',
+    globalDir:      path.join(home, '.claude'),
+    localDir:       '.claude',
+    commandsSubdir: 'commands/mindforge',
+    entryFile:      'CLAUDE.md',
+    commandFormat:  'slash',
+    supportsSlash:  true,
+    detect:         () => !!(process.env.CLAUDE_CODE || fs.existsSync('.claude/CLAUDE.md')),
+    description:    'Anthropic Claude Code (claude.ai/code)',
+    installDocs:    'https://docs.anthropic.com/claude-code',
+  },
+
+  // ── Antigravity ──────────────────────────────────────────────────────────────
+  antigravity: {
+    id:             'antigravity',
+    displayName:    'Antigravity',
+    globalDir:      path.join(home, '.antigravity'),
+    localDir:       '.agent',
+    commandsSubdir: 'mindforge',
+    entryFile:      'CLAUDE.md',
+    commandFormat:  'slash',
+    supportsSlash:  true,
+    detect:         () => fs.existsSync('.agent/CLAUDE.md'),
+    description:    'Antigravity AI coding agent',
+    installDocs:    'https://antigravity.dev',
+  },
+
+  // ── Cursor ───────────────────────────────────────────────────────────────────
+  cursor: {
+    id:             'cursor',
+    displayName:    'Cursor',
+    globalDir:      path.join(home, '.cursor', 'rules'),
+    localDir:       '.cursor/rules',
+    commandsSubdir: 'mindforge',
+    entryFile:      '.cursorrules',
+    commandFormat:  'file',          // Cursor reads command files as context rules
+    supportsSlash:  false,           // Cursor uses @-rules, not /slash commands
+    detect:         () => fs.existsSync('.cursor') || !!(process.env.CURSOR_TRACE_ID),
+    description:    'Cursor AI code editor (cursor.sh)',
+    installDocs:    'https://cursor.sh/docs',
+    notes:          'Commands are installed as Cursor rules in .cursor/rules/mindforge/. ' +
+                    'Use @mindforge-[command] in Cursor chat to reference them.',
+  },
+
+  // ── OpenCode ─────────────────────────────────────────────────────────────────
+  opencode: {
+    id:             'opencode',
+    displayName:    'OpenCode',
+    globalDir:      path.join(home, '.config', 'opencode'),
+    localDir:       '.opencode',
+    commandsSubdir: 'mindforge',
+    entryFile:      'CLAUDE.md',     // OpenCode uses CLAUDE.md as its entry point
+    commandFormat:  'slash',
+    supportsSlash:  true,
+    detect:         () => fs.existsSync('.opencode') || !!(process.env.OPENCODE_SESSION),
+    description:    'OpenCode open-source coding agent (opencode.ai)',
+    installDocs:    'https://opencode.ai/docs',
+  },
+
+  // ── Gemini CLI ───────────────────────────────────────────────────────────────
+  gemini: {
+    id:             'gemini',
+    displayName:    'Gemini CLI',
+    globalDir:      path.join(home, '.gemini'),
+    localDir:       '.gemini',
+    commandsSubdir: 'mindforge',
+    entryFile:      'GEMINI.md',     // Gemini CLI reads GEMINI.md as its instructions
+    commandFormat:  'slash',
+    supportsSlash:  true,
+    detect:         () => fs.existsSync('.gemini/GEMINI.md') || !!(process.env.GOOGLE_CLOUD_PROJECT && process.env.GEMINI_API_KEY),
+    description:    'Google Gemini CLI (developers.google.com/gemini-cli)',
+    installDocs:    'https://developers.google.com/gemini-cli',
+  },
+
+  // ── GitHub Copilot ───────────────────────────────────────────────────────────
+  copilot: {
+    id:             'copilot',
+    displayName:    'GitHub Copilot',
+    globalDir:      path.join(home, '.github'),
+    localDir:       '.github',
+    commandsSubdir: 'mindforge',
+    entryFile:      'copilot-instructions.md',  // Copilot reads this as custom instructions
+    commandFormat:  'include',       // Copilot uses @workspace + file references
+    supportsSlash:  false,
+    detect:         () => fs.existsSync('.github/copilot-instructions.md'),
+    description:    'GitHub Copilot (github.com/features/copilot)',
+    installDocs:    'https://docs.github.com/copilot/customizing-copilot',
+    notes:          'Commands are installed in .github/mindforge/. Reference with ' +
+                    '@workspace #file:.github/mindforge/[command].md in Copilot chat.',
+  },
+};
+
+// Runtime entry point content adapters
+// Different runtimes use different instructions file formats
+const ENTRY_CONTENT = {
+  'CLAUDE.md':               () => generateClaudeMd(),       // Claude Code / Antigravity / OpenCode
+  'GEMINI.md':               () => generateGeminiMd(),       // Gemini CLI
+  '.cursorrules':            () => generateCursorRules(),    // Cursor
+  'copilot-instructions.md': () => generateCopilotMd(),      // GitHub Copilot
+};
+
+/**
+ * Generate GEMINI.md — adapted from CLAUDE.md for Gemini CLI conventions.
+ * Key differences: uses Gemini-specific model references, GEMINI.md filename.
+ */
+function generateGeminiMd() {
+  const claudeMd = fs.readFileSync(path.join(__dirname, '..', '.claude', 'CLAUDE.md'), 'utf8');
+  return claudeMd
+    .replace(/CLAUDE\.md/g, 'GEMINI.md')
+    .replace(/Claude Code/g, 'Gemini CLI')
+    .replace(/claude-opus-4-5/g, 'gemini-2.5-pro')
+    .replace(/claude-sonnet-4-6/g, 'gemini-2.0-flash');
+}
+
+/**
+ * Generate .cursorrules — simplified rules for Cursor.
+ * Cursor rules are simpler than full CLAUDE.md — they describe project conventions.
+ */
+function generateCursorRules() {
+  return `# MindForge Project Rules for Cursor
+
+## Project Management
+This project uses MindForge for structured AI development.
+All tasks must follow the PLAN → EXECUTE → VERIFY → SHIP lifecycle.
+
+## Before implementing anything
+1. Check .planning/phases/ for the current phase plan files
+2. Read the relevant PLAN-[N]-[M].md for your task
+3. Follow the persona in the <persona> field
+
+## File scope
+Only modify files listed in the <files> field of your assigned PLAN file.
+
+## Quality gates
+Always run: [project test command] before marking any task complete.
+Never commit with failing tests.
+
+## Commands reference
+MindForge commands are in .cursor/rules/mindforge/
+Reference: @mindforge-[command-name]
+
+## Security
+Follow .mindforge/org/SECURITY.md for all security-sensitive work.
+Never commit secrets, API keys, or credentials.
+`;
+}
+
+/**
+ * Generate copilot-instructions.md — GitHub Copilot custom instructions.
+ * Copilot uses a simpler format than full CLAUDE.md.
+ */
+function generateCopilotMd() {
+  return `# MindForge Copilot Instructions
+
+This project uses MindForge v2.0.0 for structured AI-assisted development.
+
+## Development workflow
+Always plan before coding. Check .planning/phases/ for current phase plans.
+Read PLAN-[N]-[M].md files before implementing any feature.
+Run tests after every implementation.
+
+## Code standards
+Follow .mindforge/org/CONVENTIONS.md for all code style decisions.
+Follow .mindforge/org/SECURITY.md for security-sensitive code.
+
+## Available commands
+MindForge command documentation is in .github/mindforge/.
+Reference files in Copilot chat with: #file:.github/mindforge/[command].md
+
+## Testing
+Run the full test suite before marking any task complete.
+No exceptions: failing tests = incomplete task.
+`;
+}
+```
+
+### Update `bin/install.js` to support runtime selection
+
+```bash
+# Add --runtime flag to install.js
+# Usage: npx mindforge-cc@latest --runtime cursor
+# Usage: npx mindforge-cc@latest --runtime all
+# Usage: npx mindforge-cc@latest --runtime claude,gemini
+
+# The installer already handles --all flag (all runtimes)
+# Day 14 expands RUNTIMES map so --all now installs all 6
+```
+
+**Commit:**
+```bash
+git add bin/installer-core.js bin/install.js
+git commit -m "feat(v2-release): expand installer to 6 runtimes (Cursor, OpenCode, Gemini CLI, Copilot)"
+```
+
+---
+
+## TASK 2 — Add `/mindforge:new-runtime` command
+
+### `.claude/commands/mindforge/new-runtime.md`
+
+```markdown
+# MindForge v2 — New Runtime Command
+# Usage: /mindforge:new-runtime [runtime-name] [--entry-file filename] [--commands-dir path]
+# Version: v2.0.0
+
+## Purpose
+Scaffold MindForge support for a new AI coding agent/runtime that isn't
+included in the standard 6 supported runtimes.
+
+If your team uses an internal AI coding tool, a custom fork of an open-source
+agent, or a new runtime that was released after MindForge v2.0.0, this command
+generates the required files to integrate it.
+
+## What it creates
+
+```
+[runtime-local-dir]/
+├── [entry-file]              ← Agent instructions (adapted from CLAUDE.md)
+└── mindforge/
+    └── [48 command .md files] ← All MindForge commands adapted for runtime
+```
+
+## How it works
+
+### Step 1: Ask runtime details
+```
+What is the runtime name? (kebab-case, e.g., "my-agent")
+What directory does it use for instructions? (e.g., ".myagent")
+What is its main instructions file name? (e.g., "AGENT.md")
+Does it support /slash commands? (y/n)
+Where is its global config directory? (~/.myagent)
+```
+
+### Step 2: Generate adapted instructions file
+Create [local-dir]/[entry-file] by adapting .claude/CLAUDE.md:
+- Replace Claude-specific references with runtime-appropriate equivalents
+- Replace slash command syntax if the runtime doesn't support it
+- Keep all MindForge protocols (PLAN first, AUDIT logging, etc.)
+
+### Step 3: Mirror all commands
+Copy all 48 command files from .claude/commands/mindforge/ to
+[local-dir]/mindforge/ with runtime-appropriate adaptations.
+
+### Step 4: Update installer config
+Add the new runtime to MINDFORGE.md as a custom runtime config:
+```
+CUSTOM_RUNTIME_NAME=my-agent
+CUSTOM_RUNTIME_LOCAL_DIR=.myagent
+CUSTOM_RUNTIME_COMMANDS_SUBDIR=mindforge
+CUSTOM_RUNTIME_ENTRY_FILE=AGENT.md
+CUSTOM_RUNTIME_SLASH_SUPPORT=false
+```
+
+### Step 5: Verify and report
+```
+✅ New runtime "my-agent" scaffolded
+
+  Files created:
+    .myagent/AGENT.md               ← Entry point
+    .myagent/mindforge/              ← 48 command files
+
+  To activate:
+    1. Restart your agent with the new configuration
+    2. Run /mindforge:health to verify installation
+
+  Submit as a PR to add official support:
+    github.com/mindforge-dev/mindforge/pulls
+```
+
+## AUDIT entry
+```json
+{ "event": "new_runtime_scaffolded", "runtime_name": "[name]", "entry_file": "[file]", "commands_count": 48 }
+```
+```
+
+**Commit:**
+```bash
+cp .claude/commands/mindforge/new-runtime.md .agent/mindforge/new-runtime.md
+git add .claude/commands/mindforge/new-runtime.md .agent/mindforge/new-runtime.md
+git commit -m "feat(v2-release): add /mindforge:new-runtime command for custom runtime scaffolding"
+```
+
+---
+
+## TASK 3 — Write the v1.0.0→v2.0.0 Migration Script
+
+### `bin/migrations/1.0.0-to-2.0.0.js`
+
+```javascript
+/**
+ * MindForge Migration: v1.0.0 → v2.0.0
+ *
+ * Handles all schema changes introduced in Days 8-13:
+ *
+ * HANDOFF.json additions:
+ *   - autonomous_mode: { status, last_run_phase, last_run_at }
+ *   - steering_queue_path: ".planning/steering-queue.jsonl"
+ *   - knowledge_base_path: ".mindforge/memory/knowledge-base.jsonl"
+ *   - schema_version: "2.0.0" (from "1.0.0")
+ *
+ * AUDIT.jsonl backfill:
+ *   - Add model_used: "unknown" to existing entries (no model tracking in v1)
+ *   - Add node_repair: false to existing entries
+ *
+ * New files/directories created if absent:
+ *   - .mindforge/memory/ (knowledge graph)
+ *   - .planning/steering-queue.jsonl
+ *   - .planning/auto-state.json (initialized to IDLE)
+ *
+ * token-usage.jsonl:
+ *   - Existing entries get cost_usd: null (v1 had no cost tracking)
+ *   - Schema_version note added
+ *
+ * No destructive changes — all migrations are additive.
+ */
+'use strict';
+
+const fs   = require('fs');
+const path = require('path');
+
+const PLANNING_DIR = path.join(process.cwd(), '.planning');
+const MFDIR        = path.join(process.cwd(), '.mindforge');
+
+module.exports = {
+  fromVersion: '1.0.0',
+  toVersion:   '2.0.0',
+  description: 'Add v2.0.0 autonomous mode, knowledge graph, and model tracking fields',
+
+  async run(context = {}) {
+    const { dry_run = false, verbose = false } = context;
+    const log = msg => verbose && console.log(`  [1.0.0→2.0.0] ${msg}`);
+    const changes = [];
+
+    // ── 1. Update HANDOFF.json ──────────────────────────────────────────────
+    const handoffPath = path.join(PLANNING_DIR, 'HANDOFF.json');
+    if (fs.existsSync(handoffPath)) {
+      const handoff = JSON.parse(fs.readFileSync(handoffPath, 'utf8'));
+
+      let updated = false;
+
+      // Add autonomous_mode field if absent
+      if (!handoff.autonomous_mode) {
+        handoff.autonomous_mode = {
+          status:          'idle',
+          last_run_phase:  null,
+          last_run_at:     null,
+          total_runs:      0,
+        };
+        updated = true;
+        log('Added autonomous_mode to HANDOFF.json');
+        changes.push('HANDOFF.json: added autonomous_mode');
+      }
+
+      // Add steering_queue_path if absent
+      if (!handoff.steering_queue_path) {
+        handoff.steering_queue_path = '.planning/steering-queue.jsonl';
+        updated = true;
+        log('Added steering_queue_path to HANDOFF.json');
+        changes.push('HANDOFF.json: added steering_queue_path');
+      }
+
+      // Add knowledge_base_path if absent
+      if (!handoff.knowledge_base_path) {
+        handoff.knowledge_base_path = '.mindforge/memory/knowledge-base.jsonl';
+        updated = true;
+        log('Added knowledge_base_path to HANDOFF.json');
+        changes.push('HANDOFF.json: added knowledge_base_path');
+      }
+
+      // Bump schema_version
+      if (handoff.schema_version !== '2.0.0') {
+        handoff.schema_version = '2.0.0';
+        updated = true;
+        log('Updated schema_version to 2.0.0');
+        changes.push('HANDOFF.json: schema_version → 2.0.0');
+      }
+
+      if (updated && !dry_run) {
+        fs.writeFileSync(handoffPath, JSON.stringify(handoff, null, 2));
+      }
+    }
+
+    // ── 2. Backfill AUDIT.jsonl with v2 fields ───────────────────────────────
+    const auditPath = path.join(PLANNING_DIR, 'AUDIT.jsonl');
+    if (fs.existsSync(auditPath)) {
+      const lines   = fs.readFileSync(auditPath, 'utf8').split('\n').filter(Boolean);
+      let needsUpdate = false;
+      const updated = lines.map(line => {
+        try {
+          const entry = JSON.parse(line);
+          let changed = false;
+          if (entry.model_used === undefined) {
+            entry.model_used = 'unknown';  // v1 had no model tracking
+            changed = true;
+          }
+          if (entry.node_repair === undefined) {
+            entry.node_repair = false;  // v1 had no node repair
+            changed = true;
+          }
+          if (changed) needsUpdate = true;
+          return JSON.stringify(entry);
+        } catch { return line; } // Keep malformed lines as-is
+      });
+
+      if (needsUpdate && !dry_run) {
+        fs.writeFileSync(auditPath, updated.join('\n') + '\n');
+        log(`Backfilled ${lines.length} AUDIT.jsonl entries with v2 fields`);
+        changes.push(`AUDIT.jsonl: backfilled ${lines.length} entries with model_used + node_repair`);
+      }
+    }
+
+    // ── 3. Create new v2 directories if absent ───────────────────────────────
+    const dirs = [
+      path.join(MFDIR, 'memory'),
+      path.join(MFDIR, 'models'),
+      path.join(MFDIR, 'browser'),
+      path.join(MFDIR, 'dashboard'),
+    ];
+    for (const dir of dirs) {
+      if (!fs.existsSync(dir)) {
+        if (!dry_run) fs.mkdirSync(dir, { recursive: true });
+        log(`Created directory: ${path.relative(process.cwd(), dir)}`);
+        changes.push(`Created: ${path.relative(process.cwd(), dir)}/`);
+      }
+    }
+
+    // ── 4. Initialize auto-state.json if absent ──────────────────────────────
+    const autoStatePath = path.join(PLANNING_DIR, 'auto-state.json');
+    if (!fs.existsSync(autoStatePath)) {
+      const initialState = {
+        schema_version:  '2.0.0',
+        auto_mode_active: false,
+        status:          'idle',
+        phase:           null,
+        wave_current:    null,
+        wave_total:      null,
+        tasks_completed: 0,
+        tasks_total:     0,
+        node_repairs:    0,
+        escalations:     0,
+        elapsed_ms:      0,
+        current_task:    null,
+        last_commit:     null,
+        updated_at:      new Date().toISOString(),
+      };
+      if (!dry_run) fs.writeFileSync(autoStatePath, JSON.stringify(initialState, null, 2));
+      log('Created auto-state.json (initialized to idle)');
+      changes.push('Created: .planning/auto-state.json');
+    }
+
+    // ── 5. Initialize steering-queue.jsonl if absent ─────────────────────────
+    const steeringPath = path.join(PLANNING_DIR, 'steering-queue.jsonl');
+    if (!fs.existsSync(steeringPath)) {
+      if (!dry_run) fs.writeFileSync(steeringPath, '');
+      log('Created empty steering-queue.jsonl');
+      changes.push('Created: .planning/steering-queue.jsonl');
+    }
+
+    // ── 6. Backfill token-usage.jsonl with cost_usd field ────────────────────
+    const usagePath = path.join(MFDIR, 'metrics', 'token-usage.jsonl');
+    if (fs.existsSync(usagePath)) {
+      const lines   = fs.readFileSync(usagePath, 'utf8').split('\n').filter(Boolean);
+      let needsUpdate = false;
+      const updated = lines.map(line => {
+        try {
+          const entry = JSON.parse(line);
+          if (entry.cost_usd === undefined) {
+            entry.cost_usd = null; // v1 had no cost tracking
+            needsUpdate = true;
+          }
+          return JSON.stringify(entry);
+        } catch { return line; }
+      });
+      if (needsUpdate && !dry_run) {
+        fs.writeFileSync(usagePath, updated.join('\n') + '\n');
+        log(`Backfilled ${lines.length} token-usage.jsonl entries with cost_usd: null`);
+        changes.push(`token-usage.jsonl: backfilled ${lines.length} entries with cost_usd`);
+      }
+    }
+
+    return { success: true, changes, version_from: '1.0.0', version_to: '2.0.0' };
+  },
+};
+```
+
+**Also update `bin/migrations/schema-versions.js`** to register the new migration:
+
+```javascript
+// Add to the migrations registry:
+{
+  fromVersion: '1.0.0',
+  toVersion:   '2.0.0',
+  file:        './1.0.0-to-2.0.0.js',
+  description: 'Add v2.0.0 autonomous mode, knowledge graph, browser daemon, and model tracking',
+  breaking:    false, // All changes are additive
+}
+```
+
+**Commit:**
+```bash
+git add bin/migrations/1.0.0-to-2.0.0.js bin/migrations/schema-versions.js
+git commit -m "feat(v2-release): add v1.0.0→v2.0.0 migration script with additive schema changes"
+```
+
+---
+
+## TASK 4 — Update the Production Checklist to 65 Points
+
+### Update `.mindforge/production/production-checklist.md`
+
+Extend the existing 50-point checklist with Section F (15 new v2 items):
+
+```markdown
+# MindForge v2.0.0 Production Checklist
+
+**Release gate:** ALL 65 items must be checked before tagging v2.0.0.
+**Sections:** A (Installation) · B (Commands) · C (Governance) · D (Documentation) · E (Tests) · F (v2.0.0 Features)
+
+---
+
+## Section A — Installation (10 points) [unchanged from v1.0.0]
+
+| # | Check | Verification | Verifier | Date |
+|---|---|---|---|---|
+| A01 | `npx mindforge-cc@latest` completes without error | Run in empty dir | | |
+| A02 | `node tests/install.test.js` passes all assertions | Run test suite | | |
+| A03 | `.claude/CLAUDE.md` matches `.agent/CLAUDE.md` exactly | `diff .claude/CLAUDE.md .agent/CLAUDE.md` | | |
+| A04 | All 48 commands in `.claude/commands/mindforge/` | `ls .claude/commands/mindforge/ | wc -l` → 48 | | |
+| A05 | All 48 commands mirrored in `.agent/mindforge/` | `diff <(ls .claude/commands/mindforge/) <(ls .agent/mindforge/)` | | |
+| A06 | `MINDFORGE.md` is valid per schema | `node bin/validate-config.js` | | |
+| A07 | Node.js ≥ 18 gate works | Run with Node 16 → should fail with message | | |
+| A08 | `--dry-run` flag shows plan without executing | `npx mindforge-cc@latest --dry-run` | | |
+| A09 | Uninstall preserves `.planning/` and `.mindforge/` | Run uninstall, verify dirs exist | | |
+| A10 | Self-install detection works | Run inside mindforge-cc repo → no framework files copied | | |
+
+## Section B — Commands (10 points) [unchanged from v1.0.0]
+
+| # | Check | Verification | Verifier | Date |
+|---|---|---|---|---|
+| B01 | `/mindforge:help` lists all 48 commands | Run and count | | |
+| B02 | `/mindforge:init-project` creates all 8 planning files | Run in test dir, verify | | |
+| B03 | `/mindforge:plan-phase` produces valid PLAN XML | Run and validate schema | | |
+| B04 | `/mindforge:execute-phase` runs wave engine | Run with test plans, verify WAVE-REPORT | | |
+| B05 | `/mindforge:verify-phase` runs all 4 verification stages | Run and verify VERIFICATION.md | | |
+| B06 | `/mindforge:ship` creates PR with proper template | Run with --dry-run | | |
+| B07 | `/mindforge:status` reads from all data sources | Run and verify all 6 sections present | | |
+| B08 | `/mindforge:health --repair` fixes 7 known issue types | Introduce each issue, run health --repair | | |
+| B09 | `/mindforge:audit --summary` reads AUDIT.jsonl | Run and verify output format | | |
+| B10 | `/mindforge:retrospective` generates RETROSPECTIVE.md | Run with test data | | |
+
+## Section C — Governance Gates (10 points) [unchanged from v1.0.0]
+
+| # | Check | Verification | Verifier | Date |
+|---|---|---|---|---|
+| C01 | Gate 1: CRITICAL security finding blocks ship | Add test CRITICAL finding → ship should fail | | |
+| C02 | Gate 2: Failing tests block ship | Break a test → ship should fail | | |
+| C03 | Gate 3: Secret detection blocks and CANNOT be overridden | Add a fake API key → ship must block | | |
+| C04 | Gate 4: GDPR retention check is independent of skill loading | Uninstall data-privacy skill → Gate 4 still runs | | |
+| C05 | Gate 5: CVE detection works | Add `npm:lodash@4.17.15` (known CVE) → should flag | | |
+| C06 | Tier 3 change detection fires on auth file path | Touch `src/auth/login.ts` → must trigger Tier 3 | | |
+| C07 | Tier 3 change detection fires on code content | Add `jwt.sign(` anywhere → must trigger Tier 3 | | |
+| C08 | Emergency override works for EMERGENCY_APPROVERS only | Test with non-approver → blocked; with approver → allowed | | |
+| C09 | Tier 3 approval expires after 4 hours | Set expires_at in past → approval should be expired | | |
+| C10 | CI mode: Tier 3 always fails in CI regardless of settings | Set CI=true → Tier 3 must fail with clear message | | |
+
+## Section D — Documentation (10 points) [unchanged from v1.0.0]
+
+| # | Check | Verification | Verifier | Date |
+|---|---|---|---|---|
+| D01 | README.md covers installation, quick start, all features | Read and verify completeness | | |
+| D02 | `docs/reference/commands.md` documents all 48 commands | Count and verify | | |
+| D03 | `docs/security/SECURITY.md` has responsible disclosure info | Verify contact + disclosure timeline | | |
+| D04 | `docs/security/threat-model.md` covers 7 threat actors | Count threat actors | | |
+| D05 | All 41 ADRs indexed in `docs/architecture/decision-records-index.md` | Count and verify links | | |
+| D06 | CHANGELOG.md has complete entry for v2.0.0 | Read and verify all major features present | | |
+| D07 | Migration guide exists for v1.0.0 → v2.0.0 | Verify migration script works end-to-end | | |
+| D08 | Skills authoring guide is up to date for v2 | Verify quality scoring section added | | |
+| D09 | `docs/multi-model-guide.md` covers model routing | Read and verify Day 10 content | | |
+| D10 | `docs/self-building-skills-guide.md` covers learn command | Read and verify Day 13 content | | |
+
+## Section E — Test Coverage (10 points) [unchanged from v1.0.0]
+
+| # | Check | Verification | Verifier | Date |
+|---|---|---|---|---|
+| E01 | All 22 test suites pass × 1 | `node tests/[each].test.js` | | |
+| E02 | All 22 test suites pass × 2 (second run) | Re-run all | | |
+| E03 | All 22 test suites pass × 3 (third run) | Re-run all | | |
+| E04 | `node tests/e2e.test.js` completes full lifecycle | Run in fresh temp dir | | |
+| E05 | `node tests/migration.test.js` covers v1.0.0→v2.0.0 | Verify new migration included | | |
+| E06 | `node tests/release.test.js` passes all release validations | Run new release test suite | | |
+| E07 | Zero TypeScript errors in sdk/ | `cd sdk && npx tsc --noEmit` | | |
+| E08 | No secrets in any output file | `grep -r "sk-" bin/ .mindforge/ tests/` → no results | | |
+| E09 | `package.json` version is exactly `2.0.0` | `node -e "console.log(require('./package.json').version)"` | | |
+| E10 | `CHANGELOG.md` has `## [2.0.0]` entry at top | `head -5 CHANGELOG.md` | | |
+
+## Section F — v2.0.0 Features (15 points) [NEW for v2.0.0]
+
+| # | Check | Verification | Verifier | Date |
+|---|---|---|---|---|
+| F01 | `/mindforge:auto` runs a complete phase end-to-end | Run on test project — wave completes without intervention | | |
+| F02 | Node repair RETRY recovers from transient test failure | Inject flaky test → verify auto-recovery in AUDIT.jsonl | | |
+| F03 | Node repair DECOMPOSE splits an oversized plan | Create plan >80K tokens → verify decomposed sub-plans | | |
+| F04 | Node repair ESCALATE fires correctly and saves state | Hit max repair budget → verify ESCALATION.md created | | |
+| F05 | `/mindforge:steer` injects guidance at task boundary | Run auto mode → steer in second terminal → verify pickup | | |
+| F06 | Browser daemon starts, persists, and shuts down cleanly | Start → screenshot → verify PID file → stop | | |
+| F07 | `/mindforge:qa` identifies affected routes after changes | Run QA after UI changes → verify QA-REPORT | | |
+| F08 | `/mindforge:cross-review` runs with 2+ models | Run with ANTHROPIC_API_KEY + OPENAI_API_KEY → verify CROSS-REVIEW | | |
+| F09 | Model routing: Tier 3 tasks use SECURITY_MODEL | Log a Tier 3 task → verify model_used in AUDIT.jsonl | | |
+| F10 | Knowledge base captures entries from retrospective | Run /mindforge:retrospective → verify knowledge-base.jsonl | | |
+| F11 | Dashboard server starts and shows live AUDIT stream | Start /mindforge:dashboard → open browser → verify SSE feed | | |
+| F12 | `/mindforge:learn [url]` creates a valid SKILL.md | Run on a public docs URL → score must be ≥ 60 | | |
+| F13 | All 6 runtimes install correctly | Run installer with each runtime flag | | |
+| F14 | All v1.0.0 tests still pass (regression) | Run original 15 suites: install→e2e | | |
+| F15 | v1.0.0 → v2.0.0 migration completes without data loss | Run migration on v1.0.0 project → verify HANDOFF.json | | |
+
+---
+
+## Release Gate Procedure
+
+Before tagging v2.0.0:
+1. ALL 65 checklist items must be checked ✅
+2. Verifier name entered for each item
+3. Date entered for each item
+4. Final sign-off: `git tag -a v2.0.0 -m "The Autonomous Enterprise"`
+5. Push: `git push origin v2.0.0`
+6. npm publish: `npm publish --access public`
+```
+
+**Commit:**
+```bash
+git add .mindforge/production/production-checklist.md
+git commit -m "feat(v2-release): extend production checklist to 65 points with Section F v2.0.0 features"
+```
+
+---
+
+## TASK 5 — Write the Complete v2.0.0 CHANGELOG Entry
+
+### Update `CHANGELOG.md`
+
+Add the full v2.0.0 entry at the top:
+
+```markdown
+## [2.0.0] — "The Autonomous Enterprise" — 2026-03-23
+
+> The only agentic framework simultaneously enterprise-grade AND autonomously capable.
+> MindForge v2.0.0 ships everything your team needs to walk away and come back to
+> a built feature — with your team watching it happen in real time.
+
+### Major new features
+
+#### Autonomous Execution Mode (Days 8)
+**Branch:** `feat/mindforge-v2-autonomous-engine`
+- `/mindforge:auto [--phase N] [--milestone M] [--timeout T]` — walk-away execution mode
+- `/mindforge:steer [instruction] [--priority normal|urgent|stop]` — mid-execution guidance
+- Node repair operator: RETRY → DECOMPOSE → PRUNE → ESCALATE hierarchy
+  - RETRY: injects error context and retries with fresh context (max 1 attempt)
+  - DECOMPOSE: splits multi-domain plans into parallel sub-plans
+  - PRUNE: defers non-critical tasks to DEFERRED-ITEMS.md
+  - ESCALATE: saves state and notifies human (Tier 3 changes: immediate)
+- Stuck detection: S01 (file churn), S02 (time overrun), S03 (identical errors),
+  S04 (context explosion), S05 (cascade failures) — all with automatic recovery
+- Headless CI mode: exit 0 on timeout (state preserved), exit 1 on gate failures
+- `bin/autonomous/` — headless.js, repair-operator.js, stuck-monitor.js, steer.js
+- `tests/autonomous.test.js` (16th suite)
+
+#### Persistent Browser Runtime + Visual QA (Day 9)
+**Branch:** `feat/mindforge-v2-browser-runtime`
+- Long-lived Chromium daemon at localhost:7338 (Playwright-backed, 16 HTTP endpoints)
+- `/mindforge:browse` — navigate, click, type, screenshot, evaluate in live browser
+- `/mindforge:qa` — systematic visual QA against affected routes after execution
+- `<verify-visual>` XML tag in PLAN files — 13 directives for UI verification
+- Cookie import from Chrome, Arc, Brave, Edge (handles SQLite + Windows FILETIME epoch)
+- Named session persistence with per-session console error capture
+- Screenshot store with phase namespacing and quota enforcement
+- Auto-generated Playwright regression tests: `tests/regression/phase[N]-[slug].test.ts`
+- `bin/browser/` — 8 modules for daemon, sessions, visual verification, QA, regressions
+- `tests/browser.test.js` (17th suite)
+
+#### Multi-Model Intelligence Layer (Day 10)
+**Branch:** `feat/mindforge-v2-cross-model-review`
+- 7-model registry: Claude (Opus/Sonnet/Haiku), GPT-4o/mini, Gemini (2.5 Pro/Flash)
+- Persona-to-model routing (architect→PLANNER_MODEL, developer→EXECUTOR_MODEL, etc.)
+- Tier 3 override: auth/payment/PII tasks always use SECURITY_MODEL
+- `/mindforge:cross-review` — 3-round adversarial review (primary + adversarial + research)
+  - Consensus findings (2+ models agree) = high confidence
+  - Model-specific findings + contradiction detection
+- `/mindforge:research` — deep research with Gemini 2.5 Pro 1M context
+- `/mindforge:costs` — real-time cost tracking dashboard by model/phase/session
+- Daily budget enforcement: MODEL_COST_HARD_LIMIT_USD blocks before API calls
+- SSRF protection in research URL fetcher
+- `bin/models/` + `bin/review/` + `bin/research/` — 9 modules
+- `tests/model-routing.test.js` (18th suite)
+
+#### Persistent Knowledge Graph (Day 11)
+**Branch:** `feat/mindforge-v2-persistent-memory`
+- Append-only JSONL knowledge graph: 5 types × 4 JSONL files
+  - `architectural_decision`, `code_pattern`, `bug_pattern`, `team_preference`, `domain_knowledge`
+- TF-IDF relevance scoring with 30-second mtime-aware index cache
+- 7 automatic capture triggers: phase completion, compaction, debug sessions,
+  retrospectives, security findings, cross-review consensus, steering instructions
+- Injection guard: rejects prompt injection in knowledge content
+- Global knowledge store: `~/.mindforge/global-knowledge-base.jsonl` (cross-project)
+- Session start preloads relevant memories (🧠 Knowledge Base — N memories loaded)
+- `/mindforge:remember` — add/query/export/stats/promote/deprecate
+- `sdk/src/memory.ts` — TypeScript `MindForgeMemory` class
+- `bin/memory/` — 5 modules: store, indexer, capture, session-loader, global-sync
+- `tests/memory.test.js` (19th suite)
+
+#### Real-Time Web Observability Dashboard (Day 12)
+**Branch:** `feat/mindforge-v2-realtime-dashboard`
+- Express.js server at localhost:7339 (localhost-only, no build step needed)
+- 5-page dashboard: Activity Feed, Quality Metrics, Pending Approvals,
+  Knowledge Graph, Team Activity
+- Server-Sent Events: live AUDIT.jsonl tail, inode rotation detection
+- REST API: 10 endpoints for all project state
+- Approval UI: approve/reject Tier 2/3 from browser
+  - Tier 3 requires typing plan ID to confirm (ADR-034)
+- Quality charts: session scores, verify rates, security findings, cost trends
+- Team activity: active developers, last-seen, file conflict detection
+- `/mindforge:dashboard [--port 7339] [--open] [--stop]`
+- `bin/dashboard/` — 6 modules + single-file HTML frontend (no external deps)
+- `tests/dashboard.test.js` (20th suite)
+
+#### Self-Building Skills Platform (Day 13)
+**Branch:** `feat/mindforge-v2-self-building-skills`
+- `/mindforge:learn [url|path|npm:pkg|--session]` — documentation-to-skill converter
+  - Auto-selects Gemini 2.5 Pro (1M context) for large docs, sonnet for smaller
+  - 3-step AI pipeline: extract patterns → generate triggers → write SKILL.md
+  - SSRF protection + redirect loop prevention (max 5 hops)
+  - Pre-save injection check: rejects skills containing injection patterns
+- 7-dimension quality scorer (100 points, static analysis, < 100ms):
+  trigger_coverage(30) + mandatory_actions(25) + code_examples(20) + self_check(15)
+  + injection_safe(10) + no_placeholders(10) + version_history(10)
+  - Minimum to register: 60; minimum to publish: 80; injection = absolute FAIL
+- Auto-capture: phase completion hook detects patterns across 2+ tasks
+- `/mindforge:marketplace [search|featured|trending|install|publish]`
+  - Curated layer on top of npm registry with 12 featured skills
+- `session_quality_lift` metric: real signal of skill value
+- `bin/skills-builder/` — 6 modules
+- `tests/self-building-skills.test.js` (21st suite)
+
+#### Multi-Runtime Expansion (Day 14)
+**Branch:** `feat/mindforge-v2-release`
+- 6 supported runtimes: Claude Code, Antigravity, Cursor, OpenCode, Gemini CLI,
+  GitHub Copilot
+- Runtime-specific entry files: CLAUDE.md, GEMINI.md, .cursorrules, copilot-instructions.md
+- `/mindforge:new-runtime [name]` — scaffold support for any custom runtime
+- `bin/installer-core.js` RUNTIMES config expanded with runtime detection, display names,
+  command format flags, and install documentation links
+
+### Breaking changes from v1.0.0
+
+All changes are backwards compatible. Run `/mindforge:migrate` to update project files.
+
+- `PLAN XML`: gains optional `<verify-visual>` tag (backwards compatible)
+- `HANDOFF.json`: gains `autonomous_mode`, `steering_queue_path`, `knowledge_base_path` (optional)
+- `AUDIT.jsonl`: gains `model_used`, `node_repair` fields (optional — backfilled by migration)
+- `token-usage.jsonl`: gains `cost_usd` field per entry (optional — null for v1 entries)
+
+### Migration
+
+```bash
+# Upgrade an existing v1.0.0 project to v2.0.0
+/mindforge:migrate --from 1.0.0 --to 2.0.0
+
+# Or for a fresh install on an existing project
+npx mindforge-cc@latest --update
+```
+
+All v1.0.0 projects are fully compatible with v2.0.0. The migration is additive-only.
+
+### Architecture Decision Records
+
+v2.0.0 adds ADR-021 through ADR-041 (20 new ADRs):
+- ADR-021: Autonomy boundary — human steers intent, agent executes
+- ADR-022: Node repair hierarchy — RETRY → DECOMPOSE → PRUNE → ESCALATE
+- ADR-023: Gate 3 pre-commit timing in auto mode
+- ADR-024: Browser daemon localhost-only
+- ADR-025: `<verify-visual>` failure = `<verify>` failure
+- ADR-026: Session files gitignored — auth tokens never in git
+- ADR-027: Persona determines model; Tier 3 always overrides to SECURITY_MODEL
+- ADR-028: Adversarial second-reviewer prompt for cross-review
+- ADR-029: Gemini uses x-goog-api-key header (not URL query param)
+- ADR-030: Knowledge graph append-only — deprecate, never delete
+- ADR-031: Knowledge reinforcement on explicit acknowledgment, not on load
+- ADR-032: Global knowledge entries get 0.1 confidence penalty
+- ADR-033: Dashboard binds to localhost only
+- ADR-034: Tier 3 dashboard approvals require typing plan ID
+- ADR-035: AUDIT entry written before approval file update
+- ADR-036: Documentation is the authoritative source for skill content
+- ADR-037: Pattern frequency ≥ 2 tasks as auto-capture threshold
+- ADR-038: Minimum quality score of 60 for skill registration
+- ADR-039: 6-runtime support via unified RUNTIMES config
+- ADR-040: Migration is additive-only — no destructive schema changes in v2.0.0
+- ADR-041: v2.0.0 stable interface contract (all v2 commands, ports, schemas are stable in 2.x)
+
+### Metrics at v2.0.0
+
+| Metric | v1.0.0 | v2.0.0 |
+|---|---|---|
+| Commands | 36 | 48 |
+| Core skills | 10 | 10 + marketplace |
+| Agent personas | 8 | 9 (+ research-agent) |
+| ADRs | 20 | 41 |
+| Test suites | 15 | 22 |
+| Runtimes | 2 | 6 |
+| Network ports | 1 (SSE:7337) | 3 (SSE:7337, Browser:7338, Dashboard:7339) |
+| npm package | mindforge-cc | mindforge-cc |
+```
+
+**Commit:**
+```bash
+git add CHANGELOG.md
+git commit -m "docs(v2-release): complete v2.0.0 CHANGELOG — The Autonomous Enterprise"
+```
+
+---
+
+## TASK 6 — Write the v2.0.0 Release Test Suite
+
+### `tests/release.test.js`
+
+```javascript
+/**
+ * MindForge v2.0.0 — Release Test Suite (22nd suite)
+ * Tests all v2.0.0 release requirements:
+ * - All 48 commands exist in both runtimes
+ * - All 6 runtimes have correct config
+ * - v2.0.0 migration script is complete
+ * - Production checklist has Section F
+ * - CHANGELOG has v2.0.0 entry
+ * - Version is exactly 2.0.0
+ * - All ADRs through ADR-041 exist
+ * - All 9 v2 components present (bins)
+ * - 65-point production checklist
+ * - SDK has memory.ts module
+ *
+ * Run: node tests/release.test.js
+ */
+'use strict';
+
+const fs     = require('fs');
+const path   = require('path');
+const os     = require('os');
+const assert = require('assert');
+
+let passed = 0, failed = 0;
+
+function test(name, fn) {
+  try { fn(); console.log(`  ✅  ${name}`); passed++; }
+  catch(e) { console.error(`  ❌  ${name}\n      ${e.message}`); failed++; }
+}
+
+// ═══════════════════════════════════════════════════════════════════════
+console.log('\nMindForge v2.0.0 — Release Test Suite\n');
+
+// ── Version ───────────────────────────────────────────────────────────────────
+console.log('Version:');
+
+test('package.json version is exactly 2.0.0', () => {
+  const v = JSON.parse(fs.readFileSync('package.json', 'utf8')).version;
+  assert.strictEqual(v, '2.0.0', `Expected exactly 2.0.0, got ${v}`);
+});
+
+test('CHANGELOG.md has ## [2.0.0] entry', () => {
+  const content = fs.readFileSync('CHANGELOG.md', 'utf8');
+  assert.ok(content.includes('## [2.0.0]'), 'CHANGELOG must have [2.0.0] entry');
+  assert.ok(content.includes('Autonomous Enterprise'), 'CHANGELOG must mention "Autonomous Enterprise"');
+});
+
+// ── All 48 commands ───────────────────────────────────────────────────────────
+console.log('\nAll 48 commands:');
+
+const ALL_COMMANDS = [
+  // v1.0.0 (36)
+  'help','init-project','plan-phase','execute-phase','verify-phase','ship',
+  'next','quick','status','debug',
+  'skills','review','security-scan','map-codebase','discuss-phase',
+  'audit','milestone','complete-milestone','approve','sync-jira','sync-confluence',
+  'health','retrospective','profile-team','metrics',
+  'init-org','install-skill','publish-skill','pr-review','workspace','benchmark',
+  'update','migrate','plugins','tokens','release',
+  // v2.0.0 additions (12)
+  'auto','steer',           // Day 8
+  'browse','qa',            // Day 9
+  'cross-review','research','costs',  // Day 10
+  'remember',               // Day 11
+  'dashboard',              // Day 12
+  'learn','marketplace',    // Day 13
+  'new-runtime',            // Day 14
+];
+
+assert.strictEqual(ALL_COMMANDS.length, 48, `Expected 48 commands, counted ${ALL_COMMANDS.length}`);
+
+test('all 48 commands in .claude/commands/mindforge/', () => {
+  const missing = ALL_COMMANDS.filter(c => !fs.existsSync(`.claude/commands/mindforge/${c}.md`));
+  assert.strictEqual(missing.length, 0, `Missing: ${missing.join(', ')}`);
+});
+
+test('all 48 commands mirrored in .agent/mindforge/', () => {
+  const missing = ALL_COMMANDS.filter(c => !fs.existsSync(`.agent/mindforge/${c}.md`));
+  assert.strictEqual(missing.length, 0, `Missing in .agent/: ${missing.join(', ')}`);
+});
+
+test('no extra/unexpected command files in .claude/commands/mindforge/', () => {
+  const actual  = fs.readdirSync('.claude/commands/mindforge/').filter(f => f.endsWith('.md'));
+  const expected = new Set(ALL_COMMANDS.map(c => `${c}.md`));
+  const extra = actual.filter(f => !expected.has(f));
+  assert.strictEqual(extra.length, 0, `Unexpected commands: ${extra.join(', ')}`);
+});
+
+// ── All 6 runtimes ────────────────────────────────────────────────────────────
+console.log('\nMulti-runtime (6 runtimes):');
+
+const EXPECTED_RUNTIMES = ['claude', 'antigravity', 'cursor', 'opencode', 'gemini', 'copilot'];
+
+test('installer-core.js defines all 6 runtimes', () => {
+  const content = fs.readFileSync('bin/installer-core.js', 'utf8');
+  EXPECTED_RUNTIMES.forEach(rt => {
+    assert.ok(content.includes(`'${rt}'`) || content.includes(`"${rt}"`),
+      `Runtime "${rt}" not found in installer-core.js`);
+  });
+});
+
+test('Claude runtime has correct entry file (CLAUDE.md)', () => {
+  const content = fs.readFileSync('bin/installer-core.js', 'utf8');
+  assert.ok(content.includes("entryFile: 'CLAUDE.md'") || content.includes('entryFile: "CLAUDE.md"'),
+    'Claude runtime must use CLAUDE.md as entry file');
+});
+
+test('Gemini runtime has correct entry file (GEMINI.md)', () => {
+  const content = fs.readFileSync('bin/installer-core.js', 'utf8');
+  assert.ok(content.includes("'GEMINI.md'") || content.includes('"GEMINI.md"'),
+    'Gemini runtime must use GEMINI.md');
+});
+
+test('Cursor runtime has correct entry file (.cursorrules)', () => {
+  const content = fs.readFileSync('bin/installer-core.js', 'utf8');
+  assert.ok(content.includes("'.cursorrules'") || content.includes('".cursorrules"'),
+    'Cursor runtime must use .cursorrules');
+});
+
+test('Copilot runtime has correct entry file (copilot-instructions.md)', () => {
+  const content = fs.readFileSync('bin/installer-core.js', 'utf8');
+  assert.ok(content.includes('copilot-instructions.md'),
+    'Copilot runtime must use copilot-instructions.md');
+});
+
+test('new-runtime command exists in both runtimes', () => {
+  assert.ok(fs.existsSync('.claude/commands/mindforge/new-runtime.md'), 'Missing in .claude/');
+  assert.ok(fs.existsSync('.agent/mindforge/new-runtime.md'), 'Missing in .agent/');
+});
+
+// ── Migration script ──────────────────────────────────────────────────────────
+console.log('\nMigration (v1.0.0 → v2.0.0):');
+
+test('1.0.0-to-2.0.0.js migration script exists', () => {
+  assert.ok(fs.existsSync('bin/migrations/1.0.0-to-2.0.0.js'), 'Migration script missing');
+});
+
+test('migration script has correct fromVersion/toVersion', () => {
+  const migration = require('../bin/migrations/1.0.0-to-2.0.0.js');
+  assert.strictEqual(migration.fromVersion, '1.0.0', 'fromVersion must be 1.0.0');
+  assert.strictEqual(migration.toVersion, '2.0.0', 'toVersion must be 2.0.0');
+});
+
+test('migration adds autonomous_mode to HANDOFF.json', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mf-migrate-'));
+  const origCwd = process.cwd();
+  process.chdir(tmpDir);
+  try {
+    // Create a minimal v1.0.0 HANDOFF.json
+    fs.mkdirSync(path.join(tmpDir, '.planning'), { recursive: true });
+    fs.mkdirSync(path.join(tmpDir, '.mindforge', 'metrics'), { recursive: true });
+    fs.writeFileSync(
+      path.join(tmpDir, '.planning', 'HANDOFF.json'),
+      JSON.stringify({ schema_version: '1.0.0', next_task: 'Plan 1', _warning: 'no secrets', context_refs: [], blockers: [], decisions_needed: [] })
+    );
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'AUDIT.jsonl'), '');
+
+    const migration = require('../bin/migrations/1.0.0-to-2.0.0.js');
+    await migration.run({ verbose: false });
+
+    const updated = JSON.parse(fs.readFileSync(path.join(tmpDir, '.planning', 'HANDOFF.json'), 'utf8'));
+    assert.ok(updated.autonomous_mode, 'Migration should add autonomous_mode');
+    assert.strictEqual(updated.schema_version, '2.0.0', 'Migration should update schema_version to 2.0.0');
+    assert.ok(updated.steering_queue_path, 'Migration should add steering_queue_path');
+    assert.ok(updated.knowledge_base_path, 'Migration should add knowledge_base_path');
+  } finally { process.chdir(origCwd); fs.rmSync(tmpDir, { recursive: true, force: true }); }
+});
+
+test('migration creates auto-state.json', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mf-migrate2-'));
+  const origCwd = process.cwd();
+  process.chdir(tmpDir);
+  try {
+    fs.mkdirSync(path.join(tmpDir, '.planning'), { recursive: true });
+    fs.mkdirSync(path.join(tmpDir, '.mindforge', 'metrics'), { recursive: true });
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'HANDOFF.json'),
+      JSON.stringify({ schema_version: '1.0.0', next_task: '', _warning: 'no secrets', context_refs: [], blockers: [], decisions_needed: [] }));
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'AUDIT.jsonl'), '');
+
+    const migration = require('../bin/migrations/1.0.0-to-2.0.0.js');
+    await migration.run({ verbose: false });
+
+    assert.ok(fs.existsSync(path.join(tmpDir, '.planning', 'auto-state.json')),
+      'Migration should create auto-state.json');
+    const state = JSON.parse(fs.readFileSync(path.join(tmpDir, '.planning', 'auto-state.json'), 'utf8'));
+    assert.strictEqual(state.status, 'idle', 'auto-state.json should initialize to idle');
+  } finally { process.chdir(origCwd); fs.rmSync(tmpDir, { recursive: true, force: true }); }
+});
+
+test('migration is additive-only (no data loss)', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mf-migrate3-'));
+  const origCwd = process.cwd();
+  process.chdir(tmpDir);
+  try {
+    fs.mkdirSync(path.join(tmpDir, '.planning'), { recursive: true });
+    fs.mkdirSync(path.join(tmpDir, '.mindforge', 'metrics'), { recursive: true });
+    const originalHandoff = {
+      schema_version: '1.0.0',
+      next_task: 'Run Plan 3-04',
+      _warning: 'no secrets',
+      context_refs: ['.planning/PROJECT.md'],
+      blockers: [],
+      decisions_needed: [],
+      project: 'my-project',  // Custom field that must survive migration
+      custom_field: 'preserved',
+    };
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'HANDOFF.json'), JSON.stringify(originalHandoff));
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'AUDIT.jsonl'), '');
+
+    const migration = require('../bin/migrations/1.0.0-to-2.0.0.js');
+    await migration.run({ verbose: false });
+
+    const migrated = JSON.parse(fs.readFileSync(path.join(tmpDir, '.planning', 'HANDOFF.json'), 'utf8'));
+    // Original fields must be preserved
+    assert.strictEqual(migrated.next_task, 'Run Plan 3-04', 'next_task must be preserved');
+    assert.strictEqual(migrated.project, 'my-project', 'project field must be preserved');
+    assert.strictEqual(migrated.custom_field, 'preserved', 'Custom fields must be preserved');
+    assert.deepStrictEqual(migrated.context_refs, ['.planning/PROJECT.md'], 'context_refs must be preserved');
+  } finally { process.chdir(origCwd); fs.rmSync(tmpDir, { recursive: true, force: true }); }
+});
+
+// ── Production checklist ──────────────────────────────────────────────────────
+console.log('\nProduction checklist:');
+
+test('production-checklist.md has 65 items', () => {
+  const content = fs.readFileSync('.mindforge/production/production-checklist.md', 'utf8');
+  // Count table rows with | A/B/C/D/E/F[0-9][0-9] | pattern
+  const items = content.match(/^\| [A-F]\d{2} \|/gm) || [];
+  assert.ok(items.length >= 65, `Expected ≥65 checklist items, found ${items.length}`);
+});
+
+test('production-checklist.md has Section F (v2.0.0 features)', () => {
+  const content = fs.readFileSync('.mindforge/production/production-checklist.md', 'utf8');
+  assert.ok(content.includes('Section F'), 'Must have Section F');
+  assert.ok(content.includes('F01'), 'Must have F01 item');
+  assert.ok(content.includes('F15'), 'Must have F15 item');
+  const fItems = (content.match(/^\| F\d{2} \|/gm) || []).length;
+  assert.strictEqual(fItems, 15, `Section F must have exactly 15 items, found ${fItems}`);
+});
+
+// ── All ADRs through ADR-041 ──────────────────────────────────────────────────
+console.log('\nADRs (all 41):');
+
+test('all 41 ADRs exist in .planning/decisions/', () => {
+  if (!fs.existsSync('.planning/decisions')) {
+    console.log('    (skipping — .planning/decisions not present in dev environment)');
+    return;
+  }
+  const adrs    = fs.readdirSync('.planning/decisions').filter(f => f.startsWith('ADR-') && f.endsWith('.md'));
+  const numbers = adrs.map(f => parseInt(f.match(/ADR-(\d+)/)?.[1] || '0', 10)).filter(n => n > 0);
+  assert.ok(Math.max(...numbers) >= 41, `Max ADR number should be ≥41, got ${Math.max(...numbers)}`);
+  assert.ok(adrs.length >= 41, `Expected ≥41 ADRs, found ${adrs.length}`);
+});
+
+test('ADR decision index mentions all v2.0.0 ADRs', () => {
+  const indexPath = 'docs/architecture/decision-records-index.md';
+  if (!fs.existsSync(indexPath)) return; // Skip if docs not yet updated
+  const content = fs.readFileSync(indexPath, 'utf8');
+  assert.ok(content.includes('ADR-039') || content.includes('ADR-041'),
+    'ADR index should reference v2.0.0 ADRs');
+});
+
+// ── v2.0.0 component bins ────────────────────────────────────────────────────
+console.log('\nv2.0.0 component bins:');
+
+const V2_BINS = [
+  'bin/autonomous/headless.js',
+  'bin/autonomous/repair-operator.js',
+  'bin/autonomous/stuck-monitor.js',
+  'bin/browser/browser-daemon.js',
+  'bin/browser/visual-verify-executor.js',
+  'bin/browser/qa-engine.js',
+  'bin/models/model-client.js',
+  'bin/models/model-router.js',
+  'bin/models/cost-tracker.js',
+  'bin/review/cross-review-engine.js',
+  'bin/research/research-engine.js',
+  'bin/memory/knowledge-store.js',
+  'bin/memory/knowledge-indexer.js',
+  'bin/memory/knowledge-capture.js',
+  'bin/dashboard/server.js',
+  'bin/dashboard/sse-bridge.js',
+  'bin/dashboard/api-router.js',
+  'bin/skills-builder/skill-generator.js',
+  'bin/skills-builder/skill-scorer.js',
+  'bin/skills-builder/source-loader.js',
+];
+
+test(`all ${V2_BINS.length} v2.0.0 binary modules exist`, () => {
+  const missing = V2_BINS.filter(b => !fs.existsSync(b));
+  assert.strictEqual(missing.length, 0, `Missing v2 bins:\n  ${missing.join('\n  ')}`);
+});
+
+// ── All 22 test suites ────────────────────────────────────────────────────────
+console.log('\nAll 22 test suites:');
+
+const ALL_SUITES = [
+  // v1.0.0 (15 suites)
+  'install','wave-engine','audit','compaction','skills-platform',
+  'integrations','governance','intelligence','metrics',
+  'distribution','ci-mode','sdk','production','migration','e2e',
+  // v2.0.0 additions (7 suites)
+  'autonomous','browser','model-routing','memory','dashboard',
+  'self-building-skills','release',
+];
+
+assert.strictEqual(ALL_SUITES.length, 22);
+
+test(`all ${ALL_SUITES.length} test suites exist in tests/`, () => {
+  const missing = ALL_SUITES.filter(s => !fs.existsSync(`tests/${s}.test.js`));
+  assert.strictEqual(missing.length, 0, `Missing test suites: ${missing.join(', ')}`);
+});
+
+// ── SDK memory module ─────────────────────────────────────────────────────────
+console.log('\nSDK:');
+
+test('sdk/src/memory.ts exists (Day 11 addition)', () => {
+  assert.ok(fs.existsSync('sdk/src/memory.ts'), 'sdk/src/memory.ts must exist');
+});
+
+test('sdk/src/memory.ts exports MindForgeMemory class', () => {
+  const content = fs.readFileSync('sdk/src/memory.ts', 'utf8');
+  assert.ok(content.includes('MindForgeMemory'), 'Must export MindForgeMemory class');
+  assert.ok(content.includes('remember('), 'Must have remember() method');
+  assert.ok(content.includes('query('), 'Must have query() method');
+});
+
+// ── Network ports (all 3) ────────────────────────────────────────────────────
+console.log('\nNetwork ports (all localhost-only):');
+
+test('SDK SSE binds to 127.0.0.1:7337', () => {
+  const content = fs.readFileSync('sdk/src/events.ts', 'utf8');
+  assert.ok(content.includes('7337'), 'SDK SSE should use port 7337');
+  assert.ok(content.includes('127.0.0.1'), 'SDK SSE must bind to 127.0.0.1');
+});
+
+test('Browser daemon uses port 7338 (localhost)', () => {
+  const content = fs.readFileSync('bin/browser/browser-daemon.js', 'utf8');
+  assert.ok(content.includes('7338'), 'Browser daemon should use port 7338');
+  assert.ok(content.includes('127.0.0.1'), 'Browser daemon must bind to 127.0.0.1');
+});
+
+test('Dashboard server uses port 7339 (localhost)', () => {
+  const content = fs.readFileSync('bin/dashboard/server.js', 'utf8');
+  assert.ok(content.includes('7339'), 'Dashboard should use port 7339');
+  assert.ok(content.includes('127.0.0.1'), 'Dashboard must bind to 127.0.0.1');
+  assert.ok(!content.includes("'0.0.0.0'"), 'Dashboard must NOT bind to 0.0.0.0');
+});
+
+// ── Security: no secrets in codebase ─────────────────────────────────────────
+console.log('\nSecurity:');
+
+test('no hardcoded API keys in bin/ or .mindforge/', () => {
+  const { execSync } = require('child_process');
+  try {
+    const result = execSync(
+      'grep -r --include="*.js" --include="*.ts" --include="*.md" ' +
+      '-E "(sk-[a-zA-Z0-9]{20,}|AKIA[A-Z0-9]{16}|AIza[a-zA-Z0-9_-]{35})" ' +
+      'bin/ .mindforge/ sdk/ 2>/dev/null || echo "CLEAN"',
+      { encoding: 'utf8' }
+    );
+    const lines = result.split('\n').filter(l => l && !l.includes('CLEAN') && !l.includes('TEST_ONLY') && !l.includes('example'));
+    assert.strictEqual(lines.length, 0, `Found potential API keys:\n${lines.join('\n')}`);
+  } catch { /* grep returning non-zero = clean */ }
+});
+
+test('all server bindings checked — none on 0.0.0.0', () => {
+  const serverFiles = [
+    'bin/dashboard/server.js',
+    'bin/browser/browser-daemon.js',
+    'sdk/src/events.ts',
+  ];
+  for (const f of serverFiles) {
+    if (!fs.existsSync(f)) continue;
+    const content = fs.readFileSync(f, 'utf8');
+    assert.ok(!content.includes("'0.0.0.0'") && !content.includes('"0.0.0.0"'),
+      `${f} must NOT bind to 0.0.0.0`);
+  }
+});
+
+// ── v2.0.0 MINDFORGE.md settings ─────────────────────────────────────────────
+console.log('\nMINDFORGE.md v2.0.0 settings:');
+
+const REQUIRED_V2_SETTINGS = [
+  'PLANNER_MODEL', 'EXECUTOR_MODEL', 'SECURITY_MODEL', 'RESEARCH_MODEL',
+  'AUTO_MODE_DEFAULT_TIMEOUT_MINUTES', 'AUTO_NODE_REPAIR_BUDGET',
+  'BROWSER_PORT', 'BROWSER_HEADLESS',
+  'AUTO_CAPTURE_SKILLS', 'SKILL_QUALITY_MIN_SCORE',
+  'MODEL_COST_HARD_LIMIT_USD',
+];
+
+test('MINDFORGE.md has all required v2.0.0 settings', () => {
+  const content = fs.readFileSync('MINDFORGE.md', 'utf8');
+  const missing = REQUIRED_V2_SETTINGS.filter(s => !content.includes(s));
+  assert.strictEqual(missing.length, 0, `Missing MINDFORGE.md settings: ${missing.join(', ')}`);
+});
+
+// ── Results ───────────────────────────────────────────────────────────────────
+console.log(`\n${'═'.repeat(60)}`);
+console.log(`Release validation: ${passed} passed, ${failed} failed`);
+
+if (failed > 0) {
+  console.error(`\n❌  ${failed} release validation(s) failed.`);
+  console.error('   v2.0.0 cannot be released until all pass.\n');
+  process.exit(1);
+} else {
+  console.log('\n✅  All release validations passed.');
+  console.log('   MindForge v2.0.0 "The Autonomous Enterprise" is ready to ship.\n');
+}
+```
+
+**Commit:**
+```bash
+git add tests/release.test.js
+git commit -m "test(v2-release): add release validation suite (22nd and final test suite)"
+```
+
+---
+
+## TASK 7 — Run the Full Test Battery × 3 Runs
+
+```bash
+#!/usr/bin/env bash
+# Final pre-release test battery
+echo ""
+echo "╔══════════════════════════════════════════════════════════════╗"
+echo "║         MindForge v2.0.0 — Final Test Battery × 3          ║"
+echo "╚══════════════════════════════════════════════════════════════╝"
+
+SUITES=(install wave-engine audit compaction skills-platform \
+        integrations governance intelligence metrics \
+        distribution ci-mode sdk production migration e2e \
+        autonomous browser model-routing memory dashboard \
+        self-building-skills release)
+
+ALL_PASS=true
+
+for RUN in 1 2 3; do
+  echo ""
+  echo "  Run $RUN of 3"
+  echo "  ─────────────────────────────────────────────────────────"
+  FAIL=0
+  for s in "${SUITES[@]}"; do
+    printf "    %-35s" "${s}..."
+    node tests/${s}.test.js 2>&1 | tail -1 | grep -q "passed" && echo "✅" || { echo "❌"; ((FAIL++)); ALL_PASS=false; }
+  done
+  [ "$FAIL" -gt 0 ] && echo "  Run $RUN: $FAIL failure(s)" || echo "  Run $RUN: All 22 passed ✅"
+done
+
+echo ""
+$ALL_PASS && echo "✅ ALL 22 SUITES × 3 RUNS — RELEASE APPROVED" || { echo "❌ FAILURES — DO NOT RELEASE"; exit 1; }
+```
+
+---
+
+## TASK 8 — Update docs/architecture/decision-records-index.md
+
+```bash
+cat >> docs/architecture/decision-records-index.md << 'EOF'
+
+## v2.0.0 ADRs (Days 8-14)
+
+| ADR | Title | Day |
+|---|---|---|
+| ADR-021 | Autonomy boundary: human steers intent, agent executes implementation | 8 |
+| ADR-022 | Node repair hierarchy: RETRY before DECOMPOSE before PRUNE | 8 |
+| ADR-023 | Gate 3 runs pre-commit in auto mode (not post-wave) | 8 |
+| ADR-024 | Browser daemon localhost-only (127.0.0.1) | 9 |
+| ADR-025 | `<verify-visual>` failure = `<verify>` failure for node repair | 9 |
+| ADR-026 | Session files gitignored — auth tokens never in git | 9 |
+| ADR-027 | Persona determines model; Tier 3 always overrides to SECURITY_MODEL | 10 |
+| ADR-028 | Adversarial second-reviewer prompt for cross-review | 10 |
+| ADR-029 | Gemini uses x-goog-api-key header (not URL query param) | 10 |
+| ADR-030 | Knowledge graph is append-only (deprecate, never delete) | 11 |
+| ADR-031 | Knowledge reinforcement on explicit acknowledgment, not on load | 11 |
+| ADR-032 | Global knowledge entries carry a 0.1 confidence penalty | 11 |
+| ADR-033 | Dashboard binds to localhost only (127.0.0.1) | 12 |
+| ADR-034 | Tier 3 dashboard approvals require typing the plan ID | 12 |
+| ADR-035 | AUDIT entry written before approval file update (fail-safe ordering) | 12 |
+| ADR-036 | Documentation is the authoritative source for skill content | 13 |
+| ADR-037 | Pattern frequency ≥ 2 tasks as the auto-capture threshold | 13 |
+| ADR-038 | Minimum quality score of 60 for skill registration | 13 |
+| ADR-039 | 6-runtime support via unified RUNTIMES config map | 14 |
+| ADR-040 | v2.0.0 migration is additive-only (no destructive schema changes) | 14 |
+| ADR-041 | v2.0.0 stable interface contract (all ports, schemas, commands stable in 2.x) | 14 |
+EOF
+```
+
+**Commit:**
+```bash
+git add docs/architecture/decision-records-index.md
+git commit -m "docs(v2-release): update ADR index with all 21 v2.0.0 ADRs (021-041)"
+```
+
+---
+
+## TASK 9 — Bump to v2.0.0 and create git tag
+
+```bash
+# Final version bump
+node -e "
+  const fs = require('fs');
+  const p  = JSON.parse(fs.readFileSync('package.json','utf8'));
+  p.version = '2.0.0';
+  fs.writeFileSync('package.json', JSON.stringify(p, null, 2) + '\n');
+  console.log('Bumped to v2.0.0');
+"
+
+# Final commit for the release
+git add package.json
+git commit -m "chore(release): bump to v2.0.0 — The Autonomous Enterprise"
+
+# Tag the release
+git tag -a v2.0.0 -m "MindForge v2.0.0 — The Autonomous Enterprise
+
+48 commands · 22 test suites · 41 ADRs · 9 personas · 10 skills · 6 runtimes · 3 ports
+
+New in v2.0.0:
+- /mindforge:auto — walk-away autonomous execution with node repair
+- /mindforge:browse, /mindforge:qa — persistent browser runtime + visual QA
+- /mindforge:cross-review, /mindforge:research, /mindforge:costs — multi-model intelligence
+- /mindforge:remember — persistent knowledge graph across sessions
+- /mindforge:dashboard — real-time web UI at localhost:7339
+- /mindforge:learn, /mindforge:marketplace — self-building skills platform
+- 6 runtimes: Claude Code, Antigravity, Cursor, OpenCode, Gemini CLI, GitHub Copilot"
+
+git push origin feat/mindforge-v2-release
+git push origin v2.0.0
+```
+
+---
+
+## TASK 10 — npm pack simulation (pre-publish verification)
+
+```bash
+#!/usr/bin/env bash
+echo "npm publish simulation"
+echo "════════════════════════"
+
+# Create the tarball
+npm pack --dry-run 2>&1 | head -30
+
+# Check actual pack
+npm pack
+TARBALL=$(ls mindforge-cc-2.0.0.tgz 2>/dev/null)
+
+if [ -z "$TARBALL" ]; then
+  echo "❌ npm pack failed — tarball not created"
+  exit 1
+fi
+
+echo "✅ Tarball created: $TARBALL"
+echo "   Size: $(du -sh $TARBALL | cut -f1)"
+
+# Verify tarball contents
+echo ""
+echo "Tarball contents:"
+tar -tzf "$TARBALL" | grep -E "\.(js|ts|md|json)$" | head -40
+echo "..."
+
+# Verify critical files are included
+REQUIRED_IN_PACK=(
+  "package/bin/install.js"
+  "package/bin/installer-core.js"
+  "package/.claude/CLAUDE.md"
+  "package/.mindforge/org/SECURITY.md"
+  "package/CHANGELOG.md"
+)
+
+for f in "${REQUIRED_IN_PACK[@]}"; do
+  tar -tzf "$TARBALL" | grep -q "$f" && echo "  ✅ $f" || echo "  ❌ MISSING: $f"
+done
+
+# Verify node_modules NOT in tarball
+NODE_MODULES=$(tar -tzf "$TARBALL" | grep "node_modules" | wc -l | tr -d ' ')
+[ "$NODE_MODULES" -eq 0 ] && echo "✅ node_modules not in tarball" || echo "❌ node_modules IN tarball — check .npmignore"
+
+# Clean up
+rm -f mindforge-cc-2.0.0.tgz
+
+echo ""
+echo "✅ npm pack simulation complete — ready to publish"
+echo "   To publish: npm publish --access public"
+```
+
+**Commit:**
+```bash
+git add .
+git commit -m "chore(v2-release): final pre-publish verification — all checks passed"
+```
+
+---
+
+# ═══════════════════════════════════════════════════════════════════════
+# PART 2 — REVIEW PROMPT
+# ═══════════════════════════════════════════════════════════════════════
+
+---
+
+## DAY 14 REVIEW
+
+Activate **`architect.md` + `security-reviewer.md` + `release-manager.md`** simultaneously.
+
+Day 14 risk profile (release-specific):
+1. **Migration data loss** — v1.0.0→v2.0.0 migration must be truly additive; any destructive operation loses user data
+2. **Runtime cross-contamination** — installer copying wrong files to wrong runtime directories
+3. **Tarball completeness** — npm pack missing critical files; or including sensitive files
+4. **Version contract breakage** — v2.0.0 commands accidentally breaking v1.0.0 workflows
+5. **Test suite inter-dependency** — test suites that pass individually but fail in sequence
+6. **ADR-041 contract completeness** — the v2.0.0 stable interface contract must be precisely defined
+
+---
+
+## REVIEW PASS 1 — Migration: Data Preservation
+
+Read `bin/migrations/1.0.0-to-2.0.0.js` completely.
+
+- [ ] **`run()` does not run backup before making changes.** The v1.0.0 migration (Day 7) created a backup before applying any changes. The v2.0.0 migration writes directly to HANDOFF.json and AUDIT.jsonl without backup. If the migration is interrupted mid-way (process kill, disk full), files are partially modified. Fix: "Follow the same backup pattern as `bin/migrations/migrate.js`: create backup copy of HANDOFF.json and AUDIT.jsonl before any writes, restore on exception."
+
+- [ ] **AUDIT.jsonl backfill reads and rewrites the ENTIRE file.** For projects with large AUDIT.jsonl files (10K+ lines), this loads the entire file into memory, modifies it, and writes it back. This is an O(n) memory and disk operation. Fix: "Only backfill if needed: check the first entry in AUDIT.jsonl to see if it already has `model_used`. If it does, skip the backfill — it was already run or this is a v2.0.0 project."
+
+- [ ] **`require('../bin/migrations/1.0.0-to-2.0.0.js')` in test file may cache the module.** The test calls the migration three times across different temp directories (`process.chdir(tmpDir)`), but `require()` caches modules. The migration module reads `process.cwd()` at call time — this should work correctly. But the `require` cache means if the migration uses any module-level state (it doesn't, but document this explicitly with a comment in the migration file).
+
+---
+
+## REVIEW PASS 2 — Multi-Runtime: Installer Correctness
+
+Read the updated `bin/installer-core.js` completely.
+
+- [ ] **Cursor runtime uses `supportsSlash: false` but the commands are still `.md` files.** If Cursor doesn't support /slash commands, installing 48 command files that use `/mindforge:cmd` syntax as `@mindforge-cmd` references may confuse users. Fix: "For non-slash runtimes (cursor, copilot), add a preamble to each command file explaining how to reference it: '# To use this command in Cursor: @[filename without .md]'."
+
+- [ ] **`generateGeminiMd()` does a model string replacement on CLAUDE.md.** If the CLAUDE.md content doesn't have the expected model strings (they may not be hardcoded in CLAUDE.md — they're in MINDFORGE.md), the replacement produces the original file with no changes and no error. Fix: "Verify the replacement was meaningful: if the generated GEMINI.md is identical to CLAUDE.md after replacement, log a warning rather than silently writing a wrong file."
+
+- [ ] **The RUNTIMES config has `detect()` functions but the installer does not call them automatically.** The detection functions are only useful if the installer auto-detects which runtimes are present. If the user doesn't specify `--runtime`, the installer should call `detect()` and suggest detected runtimes. Document: "Auto-detection is provided for future use; in v2.0.0 the user explicitly specifies `--runtime`. Auto-suggest is planned for v2.1.0."
+
+---
+
+## REVIEW PASS 3 — Production Checklist: Completeness
+
+Read `.mindforge/production/production-checklist.md` completely.
+
+- [ ] **Section F item F13 says "All 6 runtimes install correctly" but doesn't specify which test to run.** The verification column says only "Run installer with each runtime flag" — this needs to be more precise. Fix: "F13 verification: `node bin/install.js --runtime cursor --dry-run && node bin/install.js --runtime gemini --dry-run && node bin/install.js --runtime copilot --dry-run`"
+
+- [ ] **The checklist has items F14 and F15 about regression and migration, but these are tested by E01-E03 (all 22 suites × 3 runs).** There is overlap between Section E and Section F for regression testing. This is intentional (belt-and-suspenders for a release gate) but should be documented to avoid confusion: "F14 and F15 are explicit human verification steps, not automated — they confirm that the human ran the tests and confirmed results."
+
+---
+
+## REVIEW PASS 4 — Release Test Suite: Coverage Gaps
+
+Read `tests/release.test.js` completely.
+
+- [ ] **Migration data preservation test uses `process.chdir()` which affects ALL subsequent tests in the same process.** If a test fails mid-chdir, the process may remain in the temp directory, causing all subsequent tests to fail with file-not-found errors. Fix: "Each migration test should use a `try/finally` block that always calls `process.chdir(origCwd)` regardless of test outcome — this is already done in the test, but add `p.cleanup()` call to ensure temp dirs are removed."
+
+- [ ] **The "no hardcoded API keys" test uses `grep` which may not be available on Windows.** The test uses `execSync('grep -r ...')`. Fix: "Add a `try/catch` around the execSync call and skip the test gracefully on Windows: `if (process.platform === 'win32') { console.log('    (skipping — grep not available on Windows)'); return; }`"
+
+- [ ] **Missing test: migration dry_run mode.** The migration has a `dry_run` parameter but no test verifies it actually prevents file writes. Add: "Test that calling `migration.run({ dry_run: true })` on a v1.0.0 HANDOFF.json leaves the file unchanged."
+
+---
+
+## REVIEW PASS 5 — ADR-041: Stable Interface Contract
+
+Read the Day 14 ADR additions.
+
+- [ ] **ADR-041 (v2.0.0 stable interface contract) is referenced but not fully defined.** Like ADR-020 (v1.0.0 contract), ADR-041 must precisely specify what is stable in v2.0.0 and cannot change without a MAJOR version bump. Ensure ADR-041 explicitly lists:
+  - All 48 command names (stable — no renames without deprecation period)
+  - All 3 network port numbers (7337, 7338, 7339 — stable)
+  - HANDOFF.json schema fields that are stable
+  - AUDIT.jsonl event types that are stable
+  - SDK public exports that are stable
+  - Plugin API version (stable)
+  - SKILL.md frontmatter format (stable)
+
+---
+
+## REVIEW PASS 6 — CHANGELOG and Documentation Completeness
+
+Read `CHANGELOG.md` v2.0.0 entry completely.
+
+- [ ] **The CHANGELOG entry doesn't mention the specific breaking changes file locations.** Users upgrading from v1.0.0 need to know exactly what changed in their `.planning/` directory. Fix: "Add to breaking changes: 'See `bin/migrations/1.0.0-to-2.0.0.js` for the complete list of schema changes. Run `/mindforge:migrate --from 1.0.0 --to 2.0.0 --dry-run` to preview changes before applying.'"
+
+- [ ] **`docs/reference/commands.md` should be updated to list all 48 commands but may still show 36.** Verify: "Check `docs/reference/commands.md` — it must list all 48 commands with v2 additions clearly marked as 'New in v2.0.0'."
+
+---
+
+## REVIEW SUMMARY TABLE
+
+```
+## Day 14 Review Summary
+
+| Category                | BLOCKING | MAJOR | MINOR | SUGGESTION |
+|-------------------------|----------|-------|-------|------------|
+| Migration Safety        |          |       |       |            |
+| Multi-Runtime Installer |          |       |       |            |
+| Production Checklist    |          |       |       |            |
+| Release Test Suite      |          |       |       |            |
+| ADR-041 Contract        |          |       |       |            |
+| CHANGELOG/Docs          |          |       |       |            |
+| **TOTAL**               |          |       |       |            |
+
+## Verdict
+[ ] ✅ APPROVED — Proceed to HARDEN section
+[ ] ⚠️  APPROVED WITH CONDITIONS
+[ ] ❌ NOT APPROVED — DO NOT RELEASE
+```
+
+---
+
+# ═══════════════════════════════════════════════════════════════════════
+# PART 3 — HARDENING PROMPT
+# ═══════════════════════════════════════════════════════════════════════
+
+---
+
+## DAY 14 HARDENING
+
+Activate **`security-reviewer.md` + `release-manager.md` + `architect.md`** simultaneously.
+
+```bash
+# Confirm all 22 suites pass before hardening
+for suite in install wave-engine audit compaction skills-platform \
+             integrations governance intelligence metrics \
+             distribution ci-mode sdk production migration e2e \
+             autonomous browser model-routing memory dashboard \
+             self-building-skills release; do
+  printf "  %-35s" "${suite}..."
+  node tests/${suite}.test.js 2>&1 | tail -1
+done
+```
+
+---
+
+## HARDEN 1 — Add backup/restore to migration script
+
+Update `bin/migrations/1.0.0-to-2.0.0.js`:
+
+```javascript
+// Add backup at the top of run():
+async function run(context = {}) {
+  const { dry_run = false, verbose = false } = context;
+  const log = msg => verbose && console.log(`  [1.0.0→2.0.0] ${msg}`);
+  const changes = [];
+
+  // ── BACKUP before any writes (consistent with migrate.js pattern) ──────────
+  const backupDir = path.join(PLANNING_DIR, '.migration-backup-2.0.0');
+  if (!dry_run) {
+    fs.mkdirSync(backupDir, { recursive: true });
+
+    const filesToBackup = [
+      path.join(PLANNING_DIR, 'HANDOFF.json'),
+      path.join(PLANNING_DIR, 'AUDIT.jsonl'),
+      path.join(MFDIR, 'metrics', 'token-usage.jsonl'),
+    ].filter(f => fs.existsSync(f));
+
+    for (const f of filesToBackup) {
+      const backupPath = path.join(backupDir, path.basename(f));
+      fs.copyFileSync(f, backupPath);
+      log(`Backed up: ${path.relative(process.cwd(), f)}`);
+    }
+  }
+
+  try {
+    // ... existing migration logic unchanged ...
+    return { success: true, changes, version_from: '1.0.0', version_to: '2.0.0' };
+  } catch (err) {
+    // Restore from backup on failure
+    if (!dry_run && fs.existsSync(backupDir)) {
+      for (const f of fs.readdirSync(backupDir)) {
+        const backupPath = path.join(backupDir, f);
+        const originalPath = f === 'HANDOFF.json' || f === 'AUDIT.jsonl'
+          ? path.join(PLANNING_DIR, f)
+          : path.join(MFDIR, 'metrics', f);
+        fs.copyFileSync(backupPath, originalPath);
+        log(`Restored from backup: ${f}`);
+      }
+    }
+    throw err;
+  }
+}
+```
+
+**Commit:**
+```bash
+git add bin/migrations/1.0.0-to-2.0.0.js
+git commit -m "harden(v2-release): add backup/restore to v2.0.0 migration script"
+```
+
+---
+
+## HARDEN 2 — Add AUDIT.jsonl smart backfill (skip if already v2)
+
+Update backfill logic in migration:
+
+```javascript
+// Replace the AUDIT.jsonl backfill section:
+const auditPath = path.join(PLANNING_DIR, 'AUDIT.jsonl');
+if (fs.existsSync(auditPath)) {
+  const content = fs.readFileSync(auditPath, 'utf8');
+  const firstLine = content.split('\n').find(l => l.trim());
+
+  if (firstLine) {
+    try {
+      const firstEntry = JSON.parse(firstLine);
+      // SMART SKIP: if first entry already has model_used, this is already v2
+      if (firstEntry.model_used !== undefined) {
+        log('AUDIT.jsonl already has v2 fields — skipping backfill');
+        // Do NOT add to changes
+      } else {
+        // Needs backfill
+        const lines   = content.split('\n').filter(Boolean);
+        const updated = lines.map(line => {
+          try {
+            const entry = JSON.parse(line);
+            if (entry.model_used  === undefined) entry.model_used  = 'unknown';
+            if (entry.node_repair === undefined) entry.node_repair = false;
+            return JSON.stringify(entry);
+          } catch { return line; }
+        });
+        if (!dry_run) fs.writeFileSync(auditPath, updated.join('\n') + '\n');
+        log(`Backfilled ${lines.length} AUDIT.jsonl entries`);
+        changes.push(`AUDIT.jsonl: backfilled ${lines.length} entries`);
+      }
+    } catch { /* malformed first line — skip backfill silently */ }
+  }
+}
+```
+
+**Commit:**
+```bash
+git add bin/migrations/1.0.0-to-2.0.0.js
+git commit -m "harden(v2-release): smart AUDIT.jsonl backfill — skip if already v2"
+```
+
+---
+
+## HARDEN 3 — Write the full ADR-041 (v2.0.0 stable interface contract)
+
+### `.planning/decisions/ADR-041-v2-stable-interface-contract.md`
+
+```markdown
+# ADR-041: v2.0.0 Stable Interface Contract
+
+**Status:** Accepted | **Date:** v2.0.0 | **Day:** 14
+
+## Context
+Like ADR-020 defined the v1.0.0 stable contract, this ADR defines exactly what
+is stable in v2.0.0 and cannot change without a MAJOR version bump (to 3.0.0).
+
+## Decision
+
+### Stable: Command names (all 48)
+The following command names are stable. They may not be renamed, removed, or
+have their core behaviour changed without a deprecation period of at least
+one minor version:
+```
+v1.0.0 (36): help, init-project, plan-phase, execute-phase, verify-phase, ship,
+next, quick, status, debug, skills, review, security-scan, map-codebase,
+discuss-phase, audit, milestone, complete-milestone, approve, sync-jira,
+sync-confluence, health, retrospective, profile-team, metrics, init-org,
+install-skill, publish-skill, pr-review, workspace, benchmark, update,
+migrate, plugins, tokens, release
+
+v2.0.0 additions (12): auto, steer, browse, qa, cross-review, research, costs,
+remember, dashboard, learn, marketplace, new-runtime
+```
+
+### Stable: Network ports
+- 7337: SDK SSE event stream — stable in 2.x
+- 7338: Browser daemon HTTP API — stable in 2.x
+- 7339: Dashboard server — stable in 2.x
+
+### Stable: HANDOFF.json schema fields
+All fields present in schema_version: "2.0.0" are stable.
+Fields may be added (minor version), but not removed or renamed (major version).
+Core stable fields: schema_version, next_task, _warning, context_refs, blockers,
+decisions_needed, phase, plan, recent_commits, recent_files, autonomous_mode,
+steering_queue_path, knowledge_base_path.
+
+### Stable: AUDIT.jsonl event types
+These event types are stable — consumers may rely on them:
+task_started, task_completed, task_failed, security_finding, quality_gate_failed,
+context_compaction, phase_completed, decision_recorded, phase_execution_started,
+phase_execution_completed, skill_learned, model_routed, approval_granted,
+approval_rejected, dashboard_started, memory_applied, cross_review_completed.
+New event types may be added (minor version).
+
+### Stable: SDK public exports (sdk/src/index.ts)
+MindForgeClient, MindForgeEventStream, MindForgeMemory, and all TypeScript
+interfaces exported from sdk/src/index.ts are stable in 2.x.
+Internal modules (not exported via index.ts) are NOT stable.
+
+### Stable: Plugin API version
+plugin_api_version: "2.0.0" is stable — plugins targeting 2.x will work
+on any 2.x version.
+
+### Stable: SKILL.md frontmatter format
+The SKILL.md frontmatter schema (name, version, status, triggers, description)
+is stable in 2.x. New frontmatter fields may be added (minor version).
+
+### Stable: PLAN XML fields
+All XML tags in PLAN files (<task>, <n>, <persona>, <files>, <context>,
+<action>, <verify>, <verify-visual>, <dependencies>, <done>) are stable.
+
+### NOT stable in 2.x (may change in minor versions)
+- Internal `bin/` module APIs (used by CLI only)
+- `.mindforge/` spec file contents (engine/governance/intelligence specs)
+- Dashboard HTML/CSS/JS structure
+- HANDOFF.json optional fields added by v2 features
+
+## Consequences
+Plugins targeting v2.0.0 will work on v2.1.0, v2.2.0, etc.
+v1.0.0 projects can upgrade to v2.x without breaking any existing workflows.
+Developers building integrations on the SDK TypeScript types can rely on them
+through all 2.x releases.
+```
+
+**Commit:**
+```bash
+git add .planning/decisions/ADR-041-v2-stable-interface-contract.md
+git commit -m "docs(adr): add ADR-041 v2.0.0 stable interface contract"
+```
+
+---
+
+## HARDEN 4 — Add migration dry_run test and grep platform guard
+
+```javascript
+// Add to tests/release.test.js:
+
+console.log('\nHardening tests:');
+
+test('migration dry_run leaves HANDOFF.json unchanged', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mf-dryrun-'));
+  const origCwd = process.cwd();
+  process.chdir(tmpDir);
+  try {
+    fs.mkdirSync(path.join(tmpDir, '.planning'), { recursive: true });
+    fs.mkdirSync(path.join(tmpDir, '.mindforge', 'metrics'), { recursive: true });
+    const originalContent = JSON.stringify({
+      schema_version: '1.0.0', next_task: 'preserved', _warning: 'no secrets', context_refs: [], blockers: [], decisions_needed: []
+    });
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'HANDOFF.json'), originalContent);
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'AUDIT.jsonl'), '');
+
+    const migration = require('../bin/migrations/1.0.0-to-2.0.0.js');
+    await migration.run({ dry_run: true, verbose: false });
+
+    // File must be unchanged
+    const afterContent = fs.readFileSync(path.join(tmpDir, '.planning', 'HANDOFF.json'), 'utf8');
+    assert.strictEqual(afterContent, originalContent, 'dry_run must not modify HANDOFF.json');
+  } finally { process.chdir(origCwd); fs.rmSync(tmpDir, { recursive: true, force: true }); }
+});
+
+test('migration smart backfill skips AUDIT.jsonl already having v2 fields', async () => {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mf-smart-'));
+  const origCwd = process.cwd();
+  process.chdir(tmpDir);
+  try {
+    fs.mkdirSync(path.join(tmpDir, '.planning'), { recursive: true });
+    fs.mkdirSync(path.join(tmpDir, '.mindforge', 'metrics'), { recursive: true });
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'HANDOFF.json'),
+      JSON.stringify({ schema_version: '1.0.0', next_task: '', _warning: 'no secrets', context_refs: [], blockers: [], decisions_needed: [] }));
+    // AUDIT.jsonl already has model_used (simulates previously migrated)
+    const alreadyV2 = JSON.stringify({ id: 'abc', timestamp: '2026-01-01T00:00:00Z', event: 'task_completed', agent: 'test', session_id: 's1', model_used: 'claude-sonnet-4-6', node_repair: false });
+    fs.writeFileSync(path.join(tmpDir, '.planning', 'AUDIT.jsonl'), alreadyV2 + '\n');
+
+    const migration = require('../bin/migrations/1.0.0-to-2.0.0.js');
+    const result = await migration.run({ verbose: false });
+
+    // Should NOT have 'AUDIT.jsonl: backfilled' in changes
+    const backfillChange = result.changes.find(c => c.includes('AUDIT.jsonl: backfilled'));
+    assert.ok(!backfillChange, 'Smart backfill should skip AUDIT.jsonl that already has v2 fields');
+  } finally { process.chdir(origCwd); fs.rmSync(tmpDir, { recursive: true, force: true }); }
+});
+```
+
+**Commit:**
+```bash
+git add tests/release.test.js
+git commit -m "test(v2-release): add dry_run and smart backfill skip tests"
+```
+
+---
+
+## HARDEN 5 — Write final ADRs (ADR-039 and ADR-040)
+
+### `.planning/decisions/ADR-039-6-runtime-support.md`
+
+```markdown
+# ADR-039: 6-runtime support via unified RUNTIMES config map
+
+**Status:** Accepted | **Date:** v2.0.0 | **Day:** 14
+
+## Context
+MindForge v1.0.0 supported Claude Code and Antigravity via hardcoded paths
+in the installer. How should additional runtimes be added?
+
+## Decision
+All runtime configurations are defined in a single RUNTIMES map in
+`bin/installer-core.js`. Each entry specifies: id, displayName, globalDir,
+localDir, commandsSubdir, entryFile, commandFormat, supportsSlash,
+detect function, description, installDocs, and optional notes.
+
+## Rationale
+A config-driven approach (vs. per-runtime code) means:
+1. New runtimes require only a new config entry + an entry file generator
+2. The installer logic is runtime-agnostic
+3. auto-detection is unified (each runtime has a `detect()` function)
+4. Tests can enumerate all runtimes from the config map
+
+## Consequences
+Adding a 7th runtime (e.g., Zed, Void) requires only a RUNTIMES entry.
+The `/mindforge:new-runtime` command lets teams add custom runtime support
+without modifying the core installer.
+```
+
+### `.planning/decisions/ADR-040-migration-additive-only.md`
+
+```markdown
+# ADR-040: v2.0.0 migration is additive-only — no destructive schema changes
+
+**Status:** Accepted | **Date:** v2.0.0 | **Day:** 14
+
+## Context
+How should v1.0.0→v2.0.0 migration handle schema changes?
+
+## Decision
+All v2.0.0 schema changes are additive. The migration adds new fields to
+existing files but never removes, renames, or modifies existing field values.
+If a field already exists with a non-null value, it is preserved unchanged.
+
+## Rationale
+Destructive migration operations are irreversible without a backup.
+Additive-only means:
+1. Migration is safe to run multiple times (idempotent for new fields)
+2. Rollback is trivial (remove the added fields, done)
+3. v1.0.0 tools can still read v2.0.0 files (graceful degradation)
+4. Custom fields in user HANDOFF.json files are never touched
+
+## Implementation
+- `autonomous_mode`, `steering_queue_path`, `knowledge_base_path` are ADDED to HANDOFF.json
+- `model_used`, `node_repair` are ADDED to existing AUDIT.jsonl entries (backfill)
+- `cost_usd` is ADDED to token-usage.jsonl entries (backfill with null)
+- Nothing is deleted, renamed, or overwritten with different semantics
+
+## Consequences
+The migration can be re-run safely on already-migrated projects (smart skip).
+Teams who customised their HANDOFF.json schema do not lose custom fields.
+All changes are verified by `tests/release.test.js` migration data preservation test.
+```
+
+**Commit:**
+```bash
+git add .planning/decisions/ADR-039*.md .planning/decisions/ADR-040*.md
+git commit -m "docs(adr): add ADR-039 6-runtime config, ADR-040 additive-only migration"
+```
+
+---
+
+## HARDEN 6 — Update docs/reference/commands.md with all 48 commands
+
+```bash
+# Verify the commands reference is current
+grep -c "^## /mindforge:" docs/reference/commands.md 2>/dev/null | tr -d ' '
+# Must be >= 48
+
+# If not, add v2.0.0 commands section:
+cat >> docs/reference/commands.md << 'EOF'
+
+---
+
+## v2.0.0 Commands (New in "The Autonomous Enterprise")
+
+### /mindforge:auto
+Walk-away autonomous execution mode. See `auto.md` for full documentation.
+
+### /mindforge:steer  
+Mid-execution guidance injection. See `steer.md`.
+
+### /mindforge:browse
+Persistent browser control with session management. See `browse.md`.
+
+### /mindforge:qa
+Visual QA against affected routes. See `qa.md`.
+
+### /mindforge:cross-review
+Multi-model adversarial code review. See `cross-review.md`.
+
+### /mindforge:research
+Deep research via Gemini 1M context window. See `research.md`.
+
+### /mindforge:costs
+Real-time cost tracking dashboard. See `costs.md`.
+
+### /mindforge:remember
+Persistent knowledge graph management. See `remember.md`.
+
+### /mindforge:dashboard
+Real-time web observability dashboard. See `dashboard.md`.
+
+### /mindforge:learn
+Convert any documentation into a reusable skill. See `learn.md`.
+
+### /mindforge:marketplace
+Community skills marketplace. See `marketplace.md`.
+
+### /mindforge:new-runtime
+Scaffold support for a new AI coding runtime. See `new-runtime.md`.
+EOF
+```
+
+**Commit:**
+```bash
+git add docs/reference/commands.md
+git commit -m "docs(v2-release): update commands reference with all 48 v2.0.0 commands"
+```
+
+---
+
+## HARDEN 7 — Final × 3 test battery and release verification
+
+```bash
+#!/usr/bin/env bash
+echo ""
+echo "╔══════════════════════════════════════════════════════════════════╗"
+echo "║        MindForge v2.0.0 — FINAL RELEASE VERIFICATION           ║"
+echo "╚══════════════════════════════════════════════════════════════════╝"
+echo ""
+
+PASS=true
+
+# ── Version check ──────────────────────────────────────────────────────────────
+V=$(node -e "console.log(require('./package.json').version)")
+[[ "${V}" == "2.0.0" ]] && echo "  Version: ${V} ✅" || { echo "  ❌ Version is ${V}, must be 2.0.0"; PASS=false; }
+
+# ── Command counts ─────────────────────────────────────────────────────────────
+CLAUDE_CMDS=$(ls .claude/commands/mindforge/*.md 2>/dev/null | wc -l | tr -d ' ')
+AGENT_CMDS=$(ls .agent/mindforge/*.md 2>/dev/null | wc -l | tr -d ' ')
+[ "$CLAUDE_CMDS" -eq 48 ] && echo "  .claude commands: ${CLAUDE_CMDS} ✅" || { echo "  ❌ .claude commands: ${CLAUDE_CMDS} (expected 48)"; PASS=false; }
+[ "$AGENT_CMDS" -eq 48 ]  && echo "  .agent commands:  ${AGENT_CMDS} ✅" || { echo "  ❌ .agent commands: ${AGENT_CMDS} (expected 48)"; PASS=false; }
+
+# ── Test battery × 3 ──────────────────────────────────────────────────────────
+SUITES=(install wave-engine audit compaction skills-platform \
+        integrations governance intelligence metrics \
+        distribution ci-mode sdk production migration e2e \
+        autonomous browser model-routing memory dashboard \
+        self-building-skills release)
+
+for RUN in 1 2 3; do
+  echo ""
+  echo "  ── Test run ${RUN}/3 ──────────────────────────────────────────────"
+  FAIL=0
+  for s in "${SUITES[@]}"; do
+    printf "    %-35s" "${s}..."
+    node tests/${s}.test.js 2>&1 | tail -1 | grep -q "passed" && echo "✅" || { echo "❌"; ((FAIL++)); PASS=false; }
+  done
+  [ "$FAIL" -gt 0 ] && echo "  ❌ Run ${RUN}: ${FAIL} failure(s)" || echo "  ✅ Run ${RUN}: All 22 passed"
+done
+
+# ── Security checks ────────────────────────────────────────────────────────────
+echo ""
+echo "  ── Security ──────────────────────────────────────────────────────"
+
+# No 0.0.0.0 bindings
+UNSAFE=$(grep -r "0\.0\.0\.0" bin/dashboard/ bin/browser/ sdk/ 2>/dev/null | grep -v "#" | grep -v "//" || true)
+[ -z "$UNSAFE" ] && echo "  Localhost-only bindings ✅" || { echo "  ❌ 0.0.0.0 binding found"; PASS=false; }
+
+# No API keys in code
+KEYS=$(grep -r --include="*.js" -E "sk-[a-zA-Z0-9]{20,}" bin/ 2>/dev/null | grep -v "TEST_ONLY\|example\|REDACTED" || true)
+[ -z "$KEYS" ] && echo "  No hardcoded API keys ✅" || { echo "  ❌ Potential API keys found"; PASS=false; }
+
+# ── CHANGELOG ─────────────────────────────────────────────────────────────────
+echo ""
+echo "  ── Documentation ─────────────────────────────────────────────────"
+head -3 CHANGELOG.md | grep -q "2.0.0" && echo "  CHANGELOG has v2.0.0 entry ✅" || { echo "  ❌ CHANGELOG missing v2.0.0"; PASS=false; }
+
+ADRS=$(ls .planning/decisions/ADR-*.md 2>/dev/null | wc -l | tr -d ' ')
+[ "$ADRS" -ge 41 ] && echo "  ADRs: ${ADRS} ✅" || { echo "  ❌ ADRs: ${ADRS} (expected ≥41)"; PASS=false; }
+
+# ── Final verdict ──────────────────────────────────────────────────────────────
+echo ""
+echo "══════════════════════════════════════════════════════════════════"
+if $PASS; then
+  echo "  ✅✅✅ ALL CHECKS PASSED — v2.0.0 RELEASE APPROVED ✅✅✅"
+  echo ""
+  echo "  Next steps:"
+  echo "    git tag -a v2.0.0 -m 'The Autonomous Enterprise'"
+  echo "    git push origin v2.0.0"
+  echo "    npm publish --access public"
+else
+  echo "  ❌ RELEASE BLOCKED — Fix failures above before releasing v2.0.0"
+  exit 1
+fi
+```
+
+**Final commit:**
+```bash
+git add .
+git commit -m "harden(v2-release): final release hardening — backup in migration, ADR-041 contract, test fixes"
+git push origin feat/mindforge-v2-release
+```
+
+---
+
+## TASK 11 — Create Final v2.0.0 State File
+
+Create `MINDFORGE-STATE-V2.md` as the authoritative context file for anyone
+continuing work after v2.0.0 ships:
+
+```bash
+# The state file is the MINDFORGE-STATE-DAY12.md from the previous session,
+# updated with Day 13 and Day 14 additions.
+# See MINDFORGE-STATE-DAY12.md for the full day-by-day history.
+# This brief state file provides the Day 14-onwards quick context.
+```
+
+```markdown
+# MINDFORGE-STATE-V2.md
+# Post-release state for v2.0.0 "The Autonomous Enterprise"
+# Use this file to start any new work after the v2.0.0 release
+
+## Version: 2.0.0 (stable release)
+## Branch: main (post-merge)
+
+## What was shipped in v2.0.0
+(see CHANGELOG.md ## [2.0.0] for complete list)
+
+## Metrics
+- Commands: 48 (36 v1.0.0 + 12 v2.0.0)
+- Test suites: 22
+- ADRs: 41
+- Personas: 9
+- Core skills: 10
+- Runtimes: 6
+- Network ports: 3 (7337/7338/7339)
+- npm: mindforge-cc@2.0.0
+
+## Known gaps for v2.1.0
+1. SDK TypeScript compilation — sdk/dist/ still not built
+2. Windows path handling — not validated on Windows
+3. Runtime auto-detection — detect() functions exist but not used in installer
+4. Lazy file reading enforcement — still advisory
+5. True parallel wave execution in interactive mode — still sequential
+6. Dashboard knowledge graph as force-directed graph — still a list view
+7. PLAN XML format — should migrate to YAML (MAJOR breaking change, deferred to v3.0.0)
+
+## v2.1.0 roadmap candidates
+- SDK TypeScript build pipeline (npx tsc in sdk/)
+- Runtime auto-detection in installer (suggest detected runtimes)
+- Dashboard WebSocket upgrade (vs current SSE for full bidirectional)
+- Skills: session_quality_lift computation (requires telemetry system)
+- /mindforge:auto --milestone M across multiple phases
+- Windows path handling validation and fix
+```
+
+**Commit:**
+```bash
+git add MINDFORGE-STATE-V2.md
+git commit -m "docs(v2-release): add post-release v2.0.0 state file for continuation"
+git push origin feat/mindforge-v2-release
+```
+
+---
+
+## DAY 14 COMPLETE — v2.0.0 SHIPPED
+
+| Component | Status |
+|---|---|
+| Multi-runtime (6): Claude, Antigravity, Cursor, OpenCode, Gemini, Copilot | ✅ |
+| `/mindforge:new-runtime` command (48th command) | ✅ |
+| v1.0.0→v2.0.0 migration script (additive-only, with backup) | ✅ |
+| 65-point production checklist (50 + 15 new Section F items) | ✅ |
+| Complete v2.0.0 CHANGELOG | ✅ |
+| Full test battery × 3 runs — all 22 suites | ✅ |
+| `tests/release.test.js` (22nd and final suite) | ✅ |
+| ADR-039 (6-runtime config), ADR-040 (additive migration), ADR-041 (stable contract) | ✅ |
+| npm pack simulation — tarball integrity verified | ✅ |
+| git tag v2.0.0 — "The Autonomous Enterprise" | ✅ |
+| MINDFORGE-STATE-V2.md — post-release continuation file | ✅ |
+
+---
+
+## FINAL v2.0.0 METRICS
+
+| Metric | v1.0.0 | v2.0.0 | Change |
+|---|---|---|---|
+| Commands | 36 | 48 | +12 |
+| Core skills | 10 | 10 + ∞ marketplace | |
+| Agent personas | 8 | 9 | +1 (research-agent) |
+| ADRs | 20 | 41 | +21 |
+| Test suites | 15 | 22 | +7 |
+| Runtimes | 2 | 6 | +4 |
+| Network ports | 1 | 3 | +2 |
+| Binary modules | ~12 | ~40 | +28 |
+| SDK modules | 5 | 6 (+memory.ts) | +1 |
+| npm package | mindforge-cc@1.0.0 | mindforge-cc@2.0.0 | |
+
+---
+
+## THE v2.0.0 COMPETITIVE POSITION
+
+> **"The only agentic framework that is simultaneously enterprise-grade AND
+> autonomously capable. Everything else is either enterprise without autonomy,
+> or autonomous without enterprise."**
+
+| Capability | MindForge v2 | MindForge v2 | BMAD v6 | Superpowers | gstack |
+|---|---|---|---|---|---|
+| Autonomous walk-away mode | ✅✅ | ✅ | ❌ | ❌ | ❌ |
+| Node repair / stuck recovery | ✅✅ | ✅ | ❌ | ❌ | ❌ |
+| Persistent browser runtime | ✅✅ | ❌ | ❌ | ❌ | ✅ |
+| Cross-model code review | ✅✅ | ❌ | ❌ | ❌ | ✅ |
+| Enterprise governance | ✅✅✅ | ❌ | Partial | ❌ | ❌ |
+| Persistent long-term memory | ✅✅ | ❌ | ❌ | ❌ | ❌ |
+| Real-time web dashboard | ✅✅ | ❌ | ❌ | ❌ | Partial |
+| Self-building skills | ✅✅ | ❌ | ❌ | ✅ | ❌ |
+| Multi-runtime support | ✅✅ (6) | ✅ (6+) | ✅ (4+) | ✅ (5+) | ✅ (1) |
+| TypeScript SDK | ✅✅ | ❌ | ❌ | ❌ | ❌ |
+| Jira/Confluence/Slack | ✅✅ | ❌ | ❌ | ❌ | ❌ |
+
+**MindForge v2.0.0 — "The Autonomous Enterprise"**
+**Branch:** `feat/mindforge-v2-release`
+**Tag:** `v2.0.0`
+**npm:** `mindforge-cc@2.0.0`