specweave 0.23.10 → 0.23.12

package/CLAUDE.md

@@ -10,1749 +10,684 @@ For **contributors to SpecWeave itself** (not users).
 
 ## 🚨 CRITICAL SAFETY RULES
 
-### 0. Thinking-Before-Acting Discipline (META RULE!)
+### 0. Think-Then-Act Discipline (META RULE!)
 
-**ALWAYS act on your reasoning BEFORE attempting operations that will fail.**
+**NEVER run commands you know will fail.** Act on reasoning BEFORE execution.
 
-**Anti-pattern**: Running commands you know will fail, then fixing the issue.
+**Common patterns:**
+- Running code before compilation → `npm run rebuild` FIRST
+- Database queries before migrations/setup
+- File ops before directory creation
 
 ```typescript
-// ❌ WRONG: Attempt → Predictable failure → Fix
-node -e "require('./dist/file.js')"  // Fails (you saw this coming!)
-npm run rebuild                       // Then fix
+// ❌ WRONG: Attempt → Fail → Fix
+node -e "require('./dist/file.js')"  // You knew this would fail!
+npm run rebuild
 
-// ✅ CORRECT: Recognize dependency → Fix → Attempt
-npm run rebuild                       // Fix first
-node -e "require('./dist/file.js')"  // Then succeed
+// ✅ CORRECT: Fix → Attempt
+npm run rebuild
+node -e "require('./dist/file.js')"
 ```
 
-**Why**: Wastes time, creates confusion, makes execution harder to follow.
-
-**How to avoid**:
-1. **Read your own thinking process** - What dependencies did you identify?
-2. **Order operations logically** - Prerequisites BEFORE dependent operations
-3. **Catch predictable failures** - If you know it will fail, don't run it yet
-
-**Example patterns to watch for**:
-- Running code before compilation (`tsc`, `npm run build`)
-- Database queries before migrations/setup
-- API calls before authentication
-- File operations before directory creation
-
 ---
 
 ### 1. Local Development Setup
 
-**SpecWeave uses Claude Code's GitHub marketplace** for plugin management. This is the **cross-platform, simple approach** that works on macOS, Linux, and Windows.
-
-#### Quick Start (Recommended - All Platforms)
+**Standard Workflow** (Cross-platform):
 
 ```bash
-# 1. Clone and install
+# Setup
 git clone https://github.com/YOUR_USERNAME/specweave.git
-cd specweave
-npm install
-
-# 2. Build TypeScript
-npm run rebuild
-
-# 3. That's it! Claude Code auto-installs from GitHub marketplace
-# Hooks execute from: ~/.claude/plugins/marketplaces/specweave/
-```
-
-#### Development Workflow
+cd specweave && npm install && npm run rebuild
 
-**How it works**: Claude Code automatically pulls your latest code from GitHub every 5-10 seconds.
-
-```bash
-# 1. Make changes in local repo
+# Development Cycle
 vim src/core/task-parser.ts
-
-# 2. Build TypeScript changes
-npm run rebuild
-
-# 3. Test locally
-npm test
-
-# 4. Commit and push to your branch
-git add . && git commit -m "feat: new feature"
-git push origin develop
-
-# 5. Wait 5-10 seconds → Claude Code auto-updates marketplace
-
-# 6. Test in Claude Code
-/specweave:increment "test feature"
-# Your latest hooks now execute automatically!
-```
-
-**For hook-only changes** (no TypeScript build needed):
-```bash
-vim plugins/specweave/hooks/post-task-completion.sh
-git add . && git commit -m "fix: improve hook logic"
+npm run rebuild && npm test
+git add . && git commit -m "feat: feature"
 git push origin develop
-# Wait 5-10 seconds → hooks updated!
+# Wait 5-10s → Claude Code auto-updates marketplace
 ```
 
-**Key Benefits**:
-- ✅ **Cross-platform** (works on Windows, no admin privileges needed)
-- ✅ **Simple** (standard git workflow, no scripts)
-- ✅ **Reliable** (Claude Code manages updates, no symlink race conditions)
-- ✅ **Team-friendly** (everyone gets updates automatically)
+**Testing Unpushed Changes:**
+- **Option 1**: Temp branch → push → test → delete
+- **Option 2**: Fork-based (`claude plugin marketplace add github:YOUR_USERNAME/specweave`)
+- **Option 3**: Symlink (Unix-only, see `.specweave/docs/internal/advanced/symlink-dev-mode.md`)
 
-#### Testing Unpushed Changes (Advanced)
-
-**Option 1: Temporary Branch** (Recommended)
-```bash
-# Create throwaway test branch
-git checkout -b temp-test-$(date +%s)
-git add . && git commit -m "temp: testing unpushed changes"
-git push origin temp-test-1234567890
-
-# Claude Code pulls from your test branch (5-10 seconds)
-# Test your changes...
-
-# Clean up when done
-git push origin --delete temp-test-1234567890
-git checkout develop && git branch -D temp-test-1234567890
-```
-
-**Option 2: Fork-Based Development**
-```bash
-# One-time: Point Claude Code to your fork
-claude plugin marketplace remove specweave
-claude plugin marketplace add github:YOUR_USERNAME/specweave
-
-# Now push to your fork's branch
-git push origin develop
-# Claude Code pulls from YOUR fork, not upstream
-```
+---
 
-**Option 3: Symlink Mode** (Advanced, Unix-Only)
-If you need **instant updates** without pushing (e.g., rapid hook iteration):
-- See: `.specweave/docs/internal/advanced/symlink-dev-mode.md`
-- ⚠️ **Warning**: Only works on macOS/Linux, not Windows
-- ⚠️ Requires bash scripts, registry manipulation, ongoing maintenance
+### 2. Increment Folder Structure
 
-#### NPM Testing Mode (End-User Experience)
+**ONLY 4 files in `.specweave/increments/####/` root**: `spec.md`, `plan.md`, `tasks.md`, `metadata.json`
 
-To test the published npm package as end-users experience it:
+**Everything else → subfolders**: `reports/`, `scripts/`, `logs/`
 
 ```bash
-# 1. Install global npm package
-npm install -g specweave
-
-# 2. Test in fresh directory
-cd /tmp && mkdir test-project && cd test-project
-specweave init .
-# ... test end-user workflows ...
-
-# 3. Resume development
-cd /path/to/specweave/repo
-# Claude Code automatically switches back to marketplace mode
+# ❌ WRONG: .specweave/increments/0046/analysis-report.md
+# ✅ CORRECT: .specweave/increments/0046/reports/analysis-report.md
 ```
 
-### 2. NEVER Pollute Increment Folders
-
-**ONLY 4 files allowed in `.specweave/increments/####/` root**:
-- `spec.md`, `plan.md`, `tasks.md`, `metadata.json`
-
-**Everything else → subfolders**:
-- `reports/` - Analysis, completion reports, validation
-- `scripts/` - Helper scripts, automation
-- `logs/` - Execution logs, debug output
-
-**Examples**:
-```bash
-# ❌ WRONG
-.specweave/increments/0046/analysis-report.md
-
-# ✅ CORRECT
-.specweave/increments/0046/reports/analysis-report.md
-
-# Fix before committing:
-mkdir -p .specweave/increments/0046/reports
-mv .specweave/increments/0046/*.md .specweave/increments/0046/reports/
-git restore .specweave/increments/0046/{spec,plan,tasks}.md
-```
+---
 
-### 3. NEVER Delete .specweave/ Directories
+### 3. Protected Directories
 
-**Protected**: `.specweave/docs/`, `.specweave/increments/`
+**Never delete**: `.specweave/docs/`, `.specweave/increments/`
 
-**Pre-commit hook blocks**:
-- `rm -rf .specweave/docs` or `rm -rf .specweave/increments`
-- Deletion of 50+ files at once in these directories
+**Pre-commit hook blocks**: Deletion of 50+ files, `rm -rf` on protected dirs
 
-**If accidental deletion**:
-```bash
-git restore .specweave/
-```
+**Recovery**: `git restore .specweave/`
 
-### 4. Test Cleanup Safety (MANDATORY)
+---
 
-**Danger**: `rm -rf` with wrong `pwd` = delete entire test suite.
+### 4. Test Cleanup Safety
 
-**REQUIRED steps before any `rm -rf`**:
+**REQUIRED before `rm -rf`**:
 1. Verify `pwd` (MUST be project root)
 2. Dry-run with `-print` (NO deletion)
-3. Count files/directories to delete
-4. Manual confirmation (type "DELETE")
-5. Execute deletion
-6. Verify results
-7. Run tests
+3. Count files to delete
+4. Manual confirmation
+5. Execute → Verify → Run tests
 
-**Never** use `rm -rf` without ALL safety checks above.
+---
 
 ### 5. NEVER Use `specweave init . --force`
 
-**Danger**: Deletes ALL increments and docs without backup (unless you bypass warnings).
+**Danger**: Deletes ALL increments/docs without backup
 
-**Safe alternative**:
-```bash
-specweave init .  # Interactive, never deletes without confirmation
-```
+**Use**: `specweave init .` (interactive, safe)
 
-### 6. Increment Completion Validation
+---
 
-**NEVER** mark increment "completed" without:
-1. All acceptance criteria checked (`- [x] **AC-...`)
-2. All tasks completed
-3. All tests passing
-4. Coverage target met
+### 6. Increment Completion
 
-**Always use**: `/specweave:done 0043` (validates before closing)
-**Never**: Manual `metadata.json` edit (blocked by pre-commit hook)
+**Always use**: `/specweave:done 0043` (validates ACs, tasks, tests, coverage)
 
-### 7. Source of Truth: tasks.md and spec.md (CRITICAL!)
+**Never**: Manual `metadata.json` edit (blocked by pre-commit hook)
 
-**THE MOST CRITICAL RULE**: Internal TODO lists are ONLY for tracking work during execution. The SOURCE OF TRUTH is ALWAYS:
-1. **tasks.md** - Task completion status (`[x]` checkboxes)
-2. **spec.md** - Acceptance criteria status (`[x]` checkboxes)
+---
 
-**MANDATORY Workflow When Using TodoWrite Tool**:
+### 7. Source of Truth: tasks.md + spec.md (CRITICAL!)
 
-```
-For EVERY task marked complete in internal TODO:
-1. ✅ Mark task as completed in internal TODO
-2. ⚠️  IMMEDIATELY update tasks.md checkbox: `[ ] pending` → `[x] completed`
-3. ⚠️  IMMEDIATELY update spec.md AC checkbox: `[ ]` → `[x]`
-4. ✅ Verify both files updated BEFORE moving to next task
-```
+**THE MOST CRITICAL RULE**: Internal TODO is ephemeral. **tasks.md + spec.md are SOURCE OF TRUTH.**
 
-**Example of CORRECT workflow**:
+**MANDATORY workflow:**
 ```typescript
-// Step 1: Complete the work
+// 1. Complete work
 await createIntegrationTest();
 
-// Step 2: Update internal TODO
+// 2. Update internal TODO
 TodoWrite([{task: "T-013", status: "completed"}]);
 
-// Step 3: IMMEDIATELY update tasks.md (NEVER skip this!)
+// 3. IMMEDIATELY update tasks.md (NEVER skip!)
 Edit("tasks.md", "**Status**: [ ] pending", "**Status**: [x] completed");
 
-// Step 4: IMMEDIATELY update spec.md ACs
+// 4. IMMEDIATELY update spec.md ACs
 Edit("spec.md", "- [ ] **AC-US1-01**", "- [x] **AC-US1-01**");
-
-// Step 5: Move to next task
-```
-
-**❌ CRITICAL ERROR - Never Do This**:
-```typescript
-// ❌ WRONG: Marking TODO as complete WITHOUT updating source files
-TodoWrite([
-  {task: "T-013", status: "completed"},  // ❌ Only internal tracking
-  {task: "T-014", status: "completed"},  // ❌ Only internal tracking
-  {task: "T-015", status: "completed"}   // ❌ Only internal tracking
-]);
-// tasks.md still shows `[ ] pending` → DESYNC!
-// This is a CRITICAL VIOLATION!
-```
-
-**Pre-Closure Validation** (MANDATORY before `/specweave:done`):
-```bash
-# 1. Verify ALL tasks marked in tasks.md
-grep "^\*\*Status\*\*:" tasks.md | grep -c "\[x\] completed"
-# Output MUST equal total_tasks in frontmatter
-
-# 2. Verify ALL ACs checked in spec.md
-grep -c "^- \[x\] \*\*AC-" spec.md
-# Output MUST equal total ACs
-
-# 3. Only then close increment
-/specweave:done 0044
-```
-
-**Why This Matters**:
-- Internal TODO is ephemeral (lost between sessions)
-- tasks.md is permanent source of truth
-- spec.md is contract with stakeholders
-- Desync = broken promises to users/team
-
-**Incident Reference**: 2025-11-19 - Increment 0044 was incorrectly closed with tasks.md showing `[ ] pending` while internal TODO showed "completed". This violated SpecWeave's core principle. See `.specweave/increments/0044-integration-testing-status-hooks/reports/INCIDENT-SOURCE-OF-TRUTH-VIOLATION.md` for full post-mortem.
-
-### 7a. Status Line Synchronization (AUTOMATIC & ENFORCED!)
-
-**CRITICAL**: Status line MUST ALWAYS reflect current task completion status.
-
-**How It Works** (Automatic):
-1. Every TodoWrite call → `post-task-completion.sh` hook fires
-2. Hook calls `update-status-line.sh` → cache updates
-3. Status line displays updated progress immediately
-
-**The ONLY Way to Complete Tasks**:
-```
-ALWAYS use TodoWrite when working on increment tasks!
-- TodoWrite triggers hooks automatically
-- Hooks update status line cache
-- Status line stays synchronized
-```
-
-**❌ NEVER Complete Tasks Without TodoWrite**:
-```typescript
-// ❌ WRONG: Direct Edit without TodoWrite
-Edit("tasks.md", "**Status**: [ ] pending", "**Status**: [x] completed");
-// Status line will NOT update! Hook never fires!
-
-// ✅ CORRECT: TodoWrite triggers automatic update
-TodoWrite([{task: "T-001", status: "in_progress"}]);
-// ... do the work ...
-TodoWrite([{task: "T-001", status: "completed"}]);
-Edit("tasks.md", "**Status**: [ ] pending", "**Status**: [x] completed");
-// Hook fires → status line updates automatically!
-```
-
-**Validation Commands**:
-```bash
-# Check if status line is in sync
-/specweave:validate-status
-
-# Manual update (emergency only)
-bash plugins/specweave/hooks/lib/update-status-line.sh
-
-# Validate all increments
-npx tsx src/core/status-line-validator.ts
-```
-
-**Detection & Recovery**:
-- Pre-commit hook validates sync (blocks commit if desync detected)
-- `/specweave:done` validates before closing (blocks if desync)
-- Automatic tests verify hook fires correctly
-
-**Why This Matters**:
-- Status line is visible to user at ALL times
-- Stale status = broken trust
-- Hooks ensure atomic updates (task complete → status updates together)
-
-**Incident Reference**: 2025-11-20 - Status line showed 21/52 tasks when actually 26/52 were complete (10% desync). Root cause: Tasks marked complete without using TodoWrite, so hooks never fired. Added validation layer and tests to prevent future occurrences.
-
-### 7b. GitHub Duplicate Prevention (CRITICAL!)
-
-**CRITICAL**: External tool items (GitHub issues, JIRA tickets, ADO work items) MUST NEVER be duplicated.
-
-**Rule**: ALWAYS use DuplicateDetector for GitHub issue creation.
-
-**Why This Matters**:
-- GitHub search has eventual consistency (2-5 second lag)
-- Race conditions can create duplicates if sync runs multiple times quickly
-- Manual search with `--limit 1` hides duplicates
-
-**The ONLY Way to Create GitHub Issues**:
-```typescript
-import { DuplicateDetector } from './duplicate-detector.js';
-
-// ✅ CORRECT: Use DuplicateDetector.createWithProtection()
-const titlePattern = `[${featureId}][${userStory.id}]`;  // e.g., "[FS-047][US-001]"
-const result = await DuplicateDetector.createWithProtection({
-  title: issueContent.title,
-  body: issueContent.body,
-  titlePattern,
-  labels: issueContent.labels,
-  milestone: milestoneTitle,
-  repo: `${owner}/${repo}`
-});
-
-// Provides automatic:
-// ✅ Phase 1 (Detection): Search for existing before create
-// ✅ Phase 2 (Verification): Count-check after creation
-// ✅ Phase 3 (Reflection): Auto-close duplicates, keep oldest
-
-// ❌ WRONG: Manual gh issue create (no duplicate protection!)
-execSync('gh issue create --title "..." --body "..."');
 ```
 
-**3-Phase Protection**:
-1. **Detection** - Search GitHub for existing issues BEFORE creating
-2. **Verification** - Count-check AFTER creation (handles eventual consistency)
-3. **Reflection** - Auto-close duplicates if detected, keep oldest
-
-**Search Limits** (CRITICAL!):
-```typescript
-// ❌ WRONG: --limit 1 hides duplicates!
-gh issue list --search "[FS-047][US-001] in:title" --limit 1
-// Returns ONLY 1 result even if 10 duplicates exist!
+**Status Line Sync (Automatic):**
+- TodoWrite → `post-task-completion.sh` hook → status line cache updates
+- **NEVER** edit tasks.md without TodoWrite (hook won't fire → desync)
+- Validation: `/specweave:validate-status`
 
-// ✅ CORRECT: --limit 50 detects duplicates
-gh issue list --search "[FS-047][US-001] in:title" --limit 50
-// Returns up to 50 results, enabling duplicate detection
-```
+**GitHub Duplicate Prevention:**
+- **ALWAYS** use `DuplicateDetector.createWithProtection()` for GitHub issues
+- 3-phase: Detection → Verification → Reflection (auto-close duplicates)
+- **NEVER** use `--limit 1` in gh searches (hides duplicates, use `--limit 50`)
+- Cleanup: `bash scripts/cleanup-duplicate-github-issues.sh --dry-run`
 
-**Cleanup Existing Duplicates**:
+**Pre-closure validation:**
 ```bash
-# Preview what will be cleaned up:
-bash scripts/cleanup-duplicate-github-issues.sh --dry-run
-
-# Actually close duplicates (keeps oldest):
-bash scripts/cleanup-duplicate-github-issues.sh
-```
-
-**Validation**:
-```bash
-# Check for duplicates:
-gh issue list --json title,createdAt --limit 100 | \
-  jq -r '.[] | .title' | sort | uniq -d
-
-# If output is empty → No duplicates! ✅
-# If output shows titles → Duplicates exist, investigate!
+grep "^\*\*Status\*\*:" tasks.md | grep -c "\[x\] completed"  # Must equal total_tasks
+grep -c "^- \[x\] \*\*AC-" spec.md                            # Must equal total ACs
+/specweave:done 0044
 ```
 
-**Why This Matters**:
-- Incident 2025-11-20: 10+ duplicate User Story issues created
-- Root Cause #1: Race condition (GitHub search eventual consistency lag)
-- Root Cause #2: `--limit 1` bug (hid duplicates in search results)
-- Root Cause #3: No post-create verification
-- **Fix**: DuplicateDetector integration (github-feature-sync.ts:149-227)
+**Incidents**:
+- 2025-11-19 (0044): Internal TODO "completed" while tasks.md showed `[ ] pending`
+- 2025-11-20: Status line 10% desync (tasks marked without TodoWrite)
+- 2025-11-20: 10+ duplicate GitHub issues (race conditions, --limit 1 bug)
 
-**Protected Files**:
-- `plugins/specweave-github/lib/github-feature-sync.ts` - Uses DuplicateDetector
-- `plugins/specweave-github/lib/duplicate-detector.ts` - 3-phase protection
-- `plugins/specweave-github/lib/github-client-v2.ts` - Fixed --limit 50
-
-**Incident Reference**: 2025-11-20 - Duplicate GitHub issues for User Stories ([SP-US-006], [SP-US-007], etc.). Root cause: Race conditions + --limit 1 bug + missing DuplicateDetector integration. Fixed by implementing 3-phase protection. See `.specweave/increments/0047-us-task-linkage/reports/DUPLICATE-GITHUB-ISSUES-ROOT-CAUSE.md`.
-
-### 8. NEVER Use `console.*` in Production Code
+---
 
-**Rule**: ALL `src/` code MUST use logger abstraction, NEVER `console.log/error/warn`.
+### 8. Logger Abstraction (NEVER `console.*`)
 
-**Why**: `console.*` pollutes test output even when errors are expected and properly handled.
+**Rule**: ALL `src/` code uses logger injection, NEVER `console.log/error/warn`
 
-**Use logger injection**:
 ```typescript
 import { Logger, consoleLogger } from '../../utils/logger.js';
 
 export class MyClass {
-  private logger: Logger;
-
   constructor(options: { logger?: Logger } = {}) {
     this.logger = options.logger ?? consoleLogger;
   }
-
-  doSomething() {
-    this.logger.log('Info message');
-    this.logger.error('Error message', error);
-  }
 }
+// In tests: new MyClass({ logger: silentLogger });
 ```
 
-**In tests, use `silentLogger`**:
+**Exception - CLI Commands**: `src/cli/commands/*.ts` may use `console.*` with comment:
 ```typescript
-import { silentLogger } from '../../src/utils/logger.js';
-
-const instance = new MyClass({ logger: silentLogger });
-```
-
-**Protection**: Code review + search before commit:
-```bash
-# Check for console.* in src/ before committing:
-git diff --cached --name-only | grep '^src/.*\.ts$' | xargs grep -n 'console\.' 2>/dev/null
+// NOTE: This CLI command is primarily user-facing output (console.log/console.error).
+// All console.* calls are legitimate user-facing exceptions as defined in CONTRIBUTING.md.
 ```
 
-**Exception: CLI Commands (User-Facing Output)**:
-
-CLI commands in `src/cli/commands/*.ts` are 99% user-facing output (colored messages, progress indicators, confirmations). These files may keep `console.*` calls with proper documentation:
-
-```typescript
-export async function myCommand(options: CommandOptions = {}) {
-  // Initialize logger (injectable for testing)
-  const logger = options.logger ?? consoleLogger;
-
-  // NOTE: This CLI command is primarily user-facing output (console.log/console.error).
-  // All console.* calls in this file are legitimate user-facing exceptions
-  // as defined in CONTRIBUTING.md (colored messages, confirmations, formatted output).
-  // Logger infrastructure available for future internal debug logs if needed.
-
-  // User-facing output (keeps console.log)
-  console.log(chalk.green('✅ Operation successful!'));
-
-  // Internal debug logs (use logger)
-  logger.log('Internal state updated');
-}
-```
-
-**Pre-commit hook allows files with documentation marker**:
-- Files with comment `"user-facing output"` or `"legitimate user-facing exceptions"` bypass console.* check
-- Pre-commit hook: `scripts/pre-commit-console-check.sh`
-- See: `.specweave/increments/0046-console-elimination/` for migration pattern
-
-### 8a. NEVER Use `fs-extra` (Native fs Migration)
-
-**Rule**: ALL code MUST use native Node.js `fs` module, NEVER `fs-extra`.
-
-**Why**: `fs-extra` causes hook failures when not installed. Native fs is faster, smaller, and always available.
-
-**Migration completed**: 2025-11-20 - All hooks converted to native fs
+---
 
-**Correct usage**:
-```typescript
-// ✅ CORRECT - Native fs (sync operations)
-import { existsSync, readFileSync, statSync } from 'fs';
+### 8a. Native fs (NEVER `fs-extra`)
 
-if (existsSync(configPath)) {
-  const content = readFileSync(configPath, 'utf-8');
-  const stats = statSync(configPath);
-}
-```
+**Rule**: Use native Node.js `fs`, NEVER `fs-extra`
 
 ```typescript
-// ✅ CORRECT - Native fs (async operations)
+// ✅ CORRECT - Native fs
+import { existsSync, readFileSync } from 'fs';
 import { promises as fs } from 'fs';
+import { mkdirpSync, writeJsonSync } from '../utils/fs-native.js';
 
-const content = await fs.readFile(configPath, 'utf-8');
-await fs.mkdir(dir, { recursive: true });
-await fs.writeFile(filePath, content, 'utf-8');
+// ❌ WRONG: import fs from 'fs-extra';
 ```
 
-```typescript
-// ✅ CORRECT - Custom utilities (from utils/fs-native.js)
-import { mkdirpSync, writeJsonSync, readJsonSync } from '../utils/fs-native.js';
-
-mkdirpSync(dir);                                  // Create directory recursively
-writeJsonSync(path, data);                        // Write JSON file
-const config = readJsonSync(configPath);         // Read JSON file
-```
-
-**Migration guide (fs-extra → native fs)**:
-```typescript
-// ❌ WRONG (fs-extra)
-import fs from 'fs-extra';
-
-fs.existsSync(path);              → existsSync(path)
-fs.readFileSync(path, 'utf-8');   → readFileSync(path, 'utf-8')
-fs.statSync(path);                → statSync(path)
-await fs.readFile(path, 'utf-8'); → await fs.readFile(path, 'utf-8')
-await fs.ensureDir(dir);          → await fs.mkdir(dir, { recursive: true })
-                                      OR mkdirpSync(dir) for sync
-await fs.writeFile(path, content); → await fs.writeFile(path, content, 'utf-8')
-fs.removeSync(path);              → removeSync(path) from fs-native.js
-await fs.copy(src, dest);         → await fs.cp(src, dest, { recursive: true })
-```
+**Migration**: `fs.existsSync → existsSync`, `fs.ensureDir → fs.mkdir(dir, {recursive: true})`, `fs.removeSync → removeSync`
 
-**Prevention**:
-1. **Pre-commit hook**: `scripts/pre-commit-fs-extra-check.sh` blocks fs-extra imports
-2. **Installation**: Hook installed automatically via `bash scripts/install-git-hooks.sh`
-3. **Legacy marker**: Add `// legacy fs-extra` comment to bypass check (temporary only)
+**Prevention**: Pre-commit hook blocks fs-extra imports (bypass: `// legacy fs-extra`)
 
-**Testing**:
-```bash
-# Verify no fs-extra in hooks
-grep -r "from 'fs-extra'" plugins/*/lib/hooks/*.js
-
-# Should return nothing (empty output)
-```
-
-**Incident Reference**: 2025-11-20 - Hook failures due to fs-extra dependency. All hooks converted to native fs. Pre-commit hook added to prevent regression.
+---
 
 ### 9. Coding Standards
 
-**Full standards**: `.specweave/docs/internal/governance/coding-standards.md`
-**Auto-discovery**: Runs during brownfield analysis or `/specweave:analyze-standards`
-
-**Critical rules (enforced during code generation)**:
+**Critical rules (enforced)**:
+1. ✅ NEVER `console.*` (use logger)
+2. ✅ ALWAYS `.js` extensions in imports
+3. ✅ Test files: `.test.ts` (NEVER `.spec.ts`)
+4. Avoid `any` type
+5. Functions < 100 lines
+6. Custom error types
+7. Comment "why" not "what"
+8. No hardcoded secrets
+9. No N+1 queries
+10. Naming: camelCase (vars), PascalCase (classes), UPPER_SNAKE_CASE (constants)
 
-1. ✅ **NEVER use `console.*`** - Use logger abstraction (already enforced above)
-2. ✅ **ALWAYS import with `.js` extensions** - Required for ESM (already enforced in Build section)
-3. ✅ **Test files MUST use `.test.ts`** suffix, never `.spec.ts` (already enforced in Testing section)
-4. **Avoid `any` type** - Use specific types or generics
-5. **Functions < 50 lines** (ideal), < 100 lines (max)
-6. **Use custom error types**, not generic Error
-7. **Comment "why" not "what"**
-8. **No hardcoded secrets** - Use environment variables
-9. **No N+1 queries** - Batch database operations
-10. **Naming**: camelCase (vars), PascalCase (classes), UPPER_SNAKE_CASE (constants)
+**Auto-discovery**: `/specweave:analyze-standards`
 
-**Auto-discovery features**:
-- **Brownfield projects**: Standards auto-detected during project analysis
-- **Manual analysis**: `/specweave:analyze-standards` - Generate comprehensive standards report
-- **Drift detection**: `/specweave:analyze-standards --drift` - Check compliance with declared standards
-- **Update standards**: `/specweave:analyze-standards --update` - Update official standards from analysis
-
-**How it works**:
-1. Scans codebase (src/**/*.ts) for naming patterns, import styles, function characteristics
-2. Parses ESLint, Prettier, TypeScript configs for enforced rules
-3. Analyzes existing CLAUDE.md, CONTRIBUTING.md for declared standards
-4. Generates evidence-based standards with confidence levels (90%+ = HIGH confidence)
-5. Detects anti-patterns: hardcoded secrets, large files (>500 lines), missing error handling
-6. Outputs to `.specweave/docs/internal/governance/coding-standards-analysis.md`
-
-**Note**: Most standards are enforced by ESLint/Prettier. This list focuses on SpecWeave-specific rules and patterns that can't be auto-fixed by linters.
+---
 
-### 10. GitHub Issue Format Policy (v0.24.0+)
+### 10. GitHub Issue Format (v0.24.0+)
 
-**CRITICAL**: ALL GitHub issues MUST use User Story-level format.
+**CRITICAL**: GitHub issues are ONLY created for User Stories, NEVER for Features!
 
-**ONLY Correct Format**:
-```
-[FS-XXX][US-YYY] User Story Title
-```
+**ONLY Correct Format**: `[FS-XXX][US-YYY] User Story Title`
 
 **Examples**:
-- ✅ `[FS-043][US-001] Status Line Shows Correct Active Increment`
-- ✅ `[FS-047][US-002] AC-Task Mapping`
-
-**Incorrect Formats** (NEVER use):
-- ❌ `[FS-047]` (Feature-only, missing US-ID)
-- ❌ `[SP-US-006]` (SP prefix, missing Feature ID)
-- ❌ `[SP-FS-047-specweave]` (SP prefix, project name)
-- ❌ `[INC-0047]` (Increment-only)
-
-**How to Create Issues**:
-```bash
-# CORRECT way (creates [FS-XXX][US-YYY] issues)
-node -e "
-const { GitHubFeatureSync } = require('./dist/plugins/specweave-github/lib/github-feature-sync.js');
-const { GitHubClientV2 } = require('./dist/plugins/specweave-github/lib/github-client-v2.js');
-
-const client = GitHubClientV2.fromRepo('anton-abyzov', 'specweave');
-const sync = new GitHubFeatureSync(client, '.specweave/docs/internal/specs', process.cwd());
-sync.syncFeatureToGitHub('FS-047').then(console.log).catch(console.error);
-"
-
-# WRONG way (deprecated, creates [FS-XXX] issues)
-/specweave:increment "feature"  # ← This no longer creates GitHub issues
+```
+✅ [FS-048][US-001] Smart Pagination During Init
+✅ [FS-048][US-002] CLI-First Defaults
+❌ [FS-048]             (Feature-only - USE MILESTONE!)
+❌ [SP-FS-048-specweave] (SP prefix - DEPRECATED!)
+❌ [FS-048-specweave]   (Project suffix - README.md ONLY, NOT GitHub!)
 ```
 
-**Why This Matters**:
-- **Features** are tracked via GitHub **Milestones** (not Issues)
-- **User Stories** are tracked via GitHub **Issues**
-- **Tasks** are tracked as **checkboxes** in User Story issue body
-
-**Architecture**:
+**Architecture** (ADR-0032 Universal Hierarchy Mapping):
 ```
-Feature (FS-047)
-  ↓ GitHub Milestone #13
-  ├─ User Story (US-001) → GitHub Issue #638: [FS-047][US-001] Title
-  ├─ User Story (US-002) → GitHub Issue #639: [FS-047][US-002] Title
-  └─ User Story (US-003) → GitHub Issue #640: [FS-047][US-003] Title
+Feature FS-048 → GitHub Milestone "FS-048: Feature Title"
+├─ User Story US-001 → Issue #XXX: [FS-048][US-001] US Title
+├─ User Story US-002 → Issue #YYY: [FS-048][US-002] US Title
+└─ User Story US-003 → Issue #ZZZ: [FS-048][US-003] US Title
 ```
 
-**Deprecated Mechanisms**:
-- ❌ `post-increment-planning.sh` hook (disabled by default)
-- ❌ `update-epic-github-issue.sh` script (deprecated)
-- ❌ `generate-epic-issue-body.ts` script (deprecated)
+**Create issues**:
+```bash
+# ✅ CORRECT: Creates User Story issues
+/specweave-github:sync FS-048
 
-These created Feature-only issues with format `[FS-XXX]` which is INCORRECT.
+# ❌ WRONG: /specweave:increment does NOT create GitHub issues
+```
 
-**Hooks Disabled**:
-- Increment-level GitHub sync in `post-increment-planning.sh` (disabled unless `SPECWEAVE_ENABLE_INCREMENT_GITHUB_SYNC=true`)
-- Epic sync in `post-task-completion.sh` (disabled unless `SPECWEAVE_ENABLE_EPIC_SYNC=true`)
+**If you see Feature-level issues** (`[FS-XXX]` without `[US-YYY]`):
+1. Close them immediately with comment explaining the violation
+2. Delete any duplicate Feature folders (e.g., FS-050 when FS-048 exists)
+3. Use `/specweave-github:sync FS-XXX` to create correct User Story issues
+4. **REPORT THE BUG** - this should never happen!
 
-**See Also**:
-- `.specweave/increments/0047-us-task-linkage/reports/GITHUB-TITLE-FORMAT-FIX-PLAN.md`
-- `plugins/specweave-github/lib/user-story-issue-builder.ts:94`
+**See**:
+- `.specweave/increments/0047-us-task-linkage/reports/FEATURE-LEVEL-GITHUB-SYNC-REMOVAL-PLAN.md`
+- `.specweave/increments/0050-*/reports/GITHUB-ISSUE-BUG-ANALYSIS-2025-11-22.md`
 
-### 11. Task Format with US-Task Linkage (v0.23.0+)
+---
 
-**CRITICAL**: ALL tasks MUST include User Story linkage fields for proper traceability.
+### 11. Task Format with US Linkage (v0.23.0+)
 
-**New Task Format** (increment 0047+):
+**Required fields**:
 ```markdown
 ### T-001: Task Title
-
-**User Story**: US-001                        ← MANDATORY: Link to parent User Story
-**Satisfies ACs**: AC-US1-01, AC-US1-02      ← MANDATORY: AC coverage mapping
+**User Story**: US-001                       ← MANDATORY
+**Satisfies ACs**: AC-US1-01, AC-US1-02     ← MANDATORY
 **Status**: [x] completed
-**Priority**: P0 (Critical)
-**Estimated Effort**: 6 hours
-
-**Description**: Detailed task description...
-
-**Implementation Steps**:
-1. Step one
-2. Step two
-
-**Test Plan**:
-- **File**: `tests/unit/component.test.ts`
-- **Tests**: TC-001, TC-002
-
-**Files Affected**:
-- `src/path/to/file.ts`
-```
-
-**Why This Matters**:
-- **Traceability**: Navigate Task ↔ User Story ↔ Acceptance Criteria ↔ Feature
-- **Living Docs Sync**: Automatic task list updates in living docs US files
-- **AC Coverage**: Validation ensures all acceptance criteria have implementing tasks
-- **Progress Tracking**: Show per-US completion (e.g., "US-001: 8/11 tasks, 73%")
-
-**Hierarchical Structure**:
-```markdown
-## User Story: US-001 - User Authentication
-
-**Linked ACs**: AC-US1-01, AC-US1-02, AC-US1-03
-**Tasks**: 11 total, 8 completed
-
-### T-001: Implement login API
-**User Story**: US-001
-**Satisfies ACs**: AC-US1-01, AC-US1-02
-...
-
-### T-002: Add JWT token generation
-**User Story**: US-001
-**Satisfies ACs**: AC-US1-02, AC-US1-03
-...
 ```
 
-**Validation Rules**:
-1. `userStory` field MUST reference valid US-ID from spec.md (format: `US-XXX`)
-2. `satisfiesACs` field MUST list valid AC-IDs from spec.md (format: `AC-USXX-YY`)
-3. AC-IDs MUST belong to correct User Story (AC-US1-XX belongs to US-001)
-4. Tasks without US linkage trigger warnings during `/specweave:validate`
-5. `/specweave:done` blocks closure if orphan tasks exist (no AC coverage)
+**Why**: Traceability (Task ↔ User Story ↔ AC ↔ Feature), Living docs auto-sync, AC coverage validation
 
-**Living Docs Auto-Sync**:
-When task marked completed in tasks.md:
-1. Hook detects change via `post-task-completion.sh`
-2. `sync-living-docs.js` parses tasks with `parseTasksWithUSLinks()`
-3. Updates living docs US file:
-   - Task list: `- [x] [T-001](link): Task title`
-   - AC checkboxes: `- [x] **AC-US1-01**` (if all tasks complete)
+**Validation**: Tasks without linkage → warnings, orphan tasks → `/specweave:done` blocks closure
 
-**Backward Compatibility**:
-- Old increments (0001-0046) work without US linkage
-- Parser supports both formats (graceful degradation)
-- Migration tool available: `npx tsx scripts/migrate-task-linkage.ts <increment-id>`
+---
 
-**See Also**:
-- Implementation: `.specweave/increments/0047-us-task-linkage/`
-- Proposal: `.specweave/increments/0046-console-elimination/reports/US-TASK-LINKAGE-PROPOSAL.md`
-- Living Docs: `.specweave/docs/public/guides/bidirectional-linking.md`
+### 12. ADR Naming Convention
 
-### 12. ADR Naming Convention (CRITICAL!)
+**Correct format**: `XXXX-decision-title.md` (4-digit, kebab-case, NO `adr-` prefix)
 
-**Architecture Decision Records (ADRs) MUST follow strict naming convention**:
+**Header**: `# ADR-XXXX: Decision Title`
 
-**Correct Format**:
-- **Filename**: `XXXX-decision-title.md` (4-digit number, kebab-case, NO `adr-` prefix)
-- **Header**: `# ADR-XXXX: Decision Title` (includes `ADR-` prefix for document clarity)
-- **Location**: `.specweave/docs/internal/architecture/adr/`
+**Location**: `.specweave/docs/internal/architecture/adr/`
 
-**Examples**:
 ```
-✅ CORRECT:
-  Filename: 0007-github-first-task-sync.md
-  Header:   # ADR-0007: GitHub-First Task-Level Synchronization
-
-✅ CORRECT:
-  Filename: 0032-universal-hierarchy-mapping.md
-  Header:   # ADR-0032: Universal Hierarchy Mapping
-
-❌ WRONG:
-  Filename: adr-0007-github-first-task-sync.md  (NO adr- prefix!)
-  Reason:   Redundant (already in /adr/ directory)
-
-❌ WRONG:
-  Filename: ADR-0007-github-first-task-sync.md  (uppercase)
-  Reason:   Filenames must be lowercase
-
-❌ WRONG:
-  Filename: 007-github-first-task-sync.md  (3-digit)
-  Reason:   MUST be 4-digit (0001-9999)
+✅ Filename: 0007-github-first-task-sync.md, Header: # ADR-0007: ...
+❌ adr-0007-github-first-task-sync.md (redundant adr- prefix)
+❌ 007-github-first-task-sync.md (3-digit)
 ```
 
-**Why This Matters**:
-- **Consistency**: All ADRs follow same pattern (sortable, predictable)
-- **Clarity**: File location indicates type (in `/adr/` directory)
-- **Tooling**: Scripts expect 4-digit format for auto-numbering
-- **Cross-references**: Links use `adr/XXXX-name.md` format
-
-**Enforcement**:
-- Architect agent validates format before creating ADRs
-- Pre-commit hook checks for `adr-XXXX-*.md` pattern and rejects
-- `/specweave:validate` command verifies ADR naming
-
-**Common Mistakes**:
-1. Adding `adr-` prefix to filename (copying from header)
-2. Using 3-digit numbers instead of 4-digit (001 vs 0001)
-3. Uppercase filenames (ADR-0007 vs 0007)
-4. Missing kebab-case (spaces or underscores instead of hyphens)
-
-**Auto-Numbering**:
-To find next available ADR number (handles duplicates correctly):
+**Auto-numbering**:
 ```bash
-ls .specweave/docs/internal/architecture/adr/*.md | \
-  grep -E '/[0-9]{4}-' | \
-  sed 's/.*\/\([0-9][0-9][0-9][0-9]\)-.*/\1/' | \
-  sort -u | \
-  tail -1 | \
+ls .specweave/docs/internal/architecture/adr/*.md | grep -E '/[0-9]{4}-' | \
+  sed 's/.*\/\([0-9][0-9][0-9][0-9]\)-.*/\1/' | sort -u | tail -1 | \
   awk '{printf "Next ADR: %04d\n", $1 + 1}'
-# Output: Next ADR: 0049
-# (Extracts unique numbers only, finds max, adds 1)
 ```
 
-### 13. Archiving Logic Anti-Patterns (CRITICAL!)
+---
 
-**NEVER use string search or substring matching for structured data matching**
+### 13. Structured Data Matching
 
-#### Anti-Pattern #1: String Search for Frontmatter Fields
+**NEVER use string search for frontmatter/IDs**:
 
 ```typescript
-// ❌ WRONG: Matches ANYWHERE in file (false positives!)
-const content = await fs.readFile('spec.md', 'utf-8');
-if (content.includes('FS-039')) {
-  // This matches:
-  // - feature_id: FS-039 ✓ (correct)
-  // - "See FS-039 for details" ✗ (FALSE POSITIVE!)
-  // - [Related](FS-039) ✗ (FALSE POSITIVE!)
-}
+// ❌ WRONG: Matches ANYWHERE (false positives!)
+content.includes('FS-039')  // Matches "See FS-039" in docs!
 
 // ✅ CORRECT: Parse frontmatter explicitly
-const featureIdMatch = content.match(/^feature_id:\s*["']?([^"'\n]+)["']?$/m);
-if (featureIdMatch && featureIdMatch[1].trim() === 'FS-039') {
-  // Only matches actual frontmatter field
-}
-```
-
-#### Anti-Pattern #2: Substring Matching for IDs
+const match = content.match(/^feature_id:\s*["']?([^"'\n]+)["']?$/m);
+if (match && match[1].trim() === 'FS-039') { /* ... */ }
 
-```typescript
-// ❌ WRONG: Substring matching (false positives!)
-const isArchived = archivedList.some(item => item.includes(searchId));
-// "0039-ultra-smart-v2".includes("0039-ultra-smart") → TRUE (WRONG!)
+// ❌ WRONG: Substring matching
+archivedList.some(item => item.includes(searchId))
 
 // ✅ CORRECT: Exact equality
-const isArchived = archivedList.some(item => item === searchId);
+archivedList.some(item => item === searchId)
 ```
 
-**Why This Matters**:
-- **Incident 2025-11-20**: String search caused 11 features to be incorrectly archived
-- Features that merely REFERENCED another feature appeared to BELONG to it
-- Substring matching confused similar IDs as identical
-
-**Prevention**:
-1. Always parse structured data (YAML frontmatter, JSON) explicitly
-2. Use exact equality (`===`) for ID matching, never `.includes()`
-3. Add comprehensive logging to show matching decisions
-4. Test with edge cases: references, links, partial matches
-
-**See Also**: `.specweave/increments/0047-us-task-linkage/reports/CRITICAL-ARCHIVING-BUGS-FIX.md`
-
-### 14. Marketplace Plugin Completeness (CRITICAL!)
-
-**NEVER add incomplete plugins to `.claude-plugin/marketplace.json`**
-
-**Rule**: ONLY plugins with complete implementation (agents/, commands/, or lib/) may be listed in marketplace.json.
-
-**Why This Matters**:
-When users install `npm i -g specweave`, Claude Code loads ALL plugins listed in marketplace.json. Incomplete plugins (skeleton-only with just skills/) cause loading errors that confuse users and damage the framework's reputation.
-
-**Incomplete Plugin Definition**:
-A plugin is considered INCOMPLETE if it ONLY has:
-- `.claude-plugin/` directory
-- `skills/` directory
-
-A plugin is COMPLETE if it has ANY of:
-- `agents/` directory (specialized Claude Code agents)
-- `commands/` directory (slash commands)
-- `lib/` directory (shared utilities, TypeScript implementations)
-
-**Validation Process** (MANDATORY before ANY marketplace.json changes):
-
-```bash
-# 1. Run validation script
-bash scripts/validate-marketplace-plugins.sh
+**Incident**: 2025-11-20 - 11 features incorrectly archived (string search false positives)
 
-# Expected output: "✅ VALIDATION PASSED!"
-# If FAILED: Remove incomplete plugins from marketplace.json
-```
-
-**Examples**:
-
-```bash
-# ✅ COMPLETE plugins (may be listed in marketplace.json):
-plugins/specweave-github/
-  ├── .claude-plugin/
-  ├── agents/          ← Has agents
-  ├── commands/        ← Has commands
-  ├── lib/             ← Has lib
-  └── skills/
-
-plugins/specweave-kafka/
-  ├── .claude-plugin/
-  ├── agents/          ← Has agents
-  ├── lib/             ← Has lib
-  └── skills/
-
-# ✗ INCOMPLETE plugins (MUST NOT be listed in marketplace.json):
-plugins/specweave-cost-optimizer/
-  ├── .claude-plugin/
-  └── skills/          ← ONLY skills!
-
-plugins/specweave-figma/
-  ├── .claude-plugin/
-  └── skills/          ← ONLY skills!
-```
-
-**Adding New Plugins**:
-
-```bash
-# 1. Create plugin with COMPLETE structure
-mkdir -p plugins/specweave-newplugin/{.claude-plugin,agents,skills}
-
-# 2. Implement at least ONE of: agents/, commands/, or lib/
-# (Don't just create empty directories!)
-
-# 3. Add to marketplace.json
-vim .claude-plugin/marketplace.json
+---
 
-# 4. VALIDATE (this is CRITICAL!)
-bash scripts/validate-marketplace-plugins.sh
+### 14. Marketplace Plugin Completeness
 
-# 5. Update bin/fix-marketplace-errors.sh
-vim bin/fix-marketplace-errors.sh
-# Add "specweave-newplugin" to plugins=(...) array
+**NEVER add incomplete plugins to marketplace.json**
 
-# 6. Commit
-git add .claude-plugin/marketplace.json bin/fix-marketplace-errors.sh
-git commit -m "feat: add specweave-newplugin to marketplace"
-```
+**Complete plugin requires**: `agents/`, `commands/`, OR `lib/` (not just `.claude-plugin/` + `skills/`)
 
-**Pre-Commit Hook** (TODO - add this):
+**MANDATORY validation**:
 ```bash
-# .git/hooks/pre-commit should run:
-bash scripts/validate-marketplace-plugins.sh
-# Block commit if validation fails
+bash scripts/validate-marketplace-plugins.sh  # Must see "✅ VALIDATION PASSED!"
 ```
 
-**Incident History**:
-- **2025-11-20**: Global npm install caused 8 incomplete plugins to fail loading, generating confusing errors for users. Root cause: marketplace.json listed skeleton plugins that weren't ready for distribution.
+**Adding new plugin**:
+1. Create with agents/commands/lib (NOT empty dirs!)
+2. Add to marketplace.json
+3. **VALIDATE** (critical!)
+4. Update `bin/fix-marketplace-errors.sh`
+5. Test: `npm pack && npm i -g ./specweave-*.tgz`
 
-**Prevention**:
-1. Run `bash scripts/validate-marketplace-plugins.sh` before EVERY marketplace.json change
-2. NEVER add plugins to marketplace.json until they have agents/, commands/, or lib/
-3. Update bin/fix-marketplace-errors.sh whenever marketplace.json changes
-4. Test global npm install (`npm pack && npm i -g ./specweave-*.tgz`) before releases
+**Incident**: 2025-11-20 - 8 incomplete plugins failed loading on global install
 
-**See Also**:
-- Validation script: `scripts/validate-marketplace-plugins.sh`
-- Fix script: `bin/fix-marketplace-errors.sh`
-
-### 15. Skills vs Agents: Understanding the Distinction (CRITICAL!)
-
-**NEVER confuse skills with agents** - They are different components with different invocation methods.
+---
 
-**The Problem**:
-Empty agent directories cause "Agent type not found" errors when someone tries to invoke them. This happened with `specweave:increment-quality-judge-v2` which existed as a skill but had an empty agent directory.
+### 15. Skills vs Agents
 
 **Key Differences**:
 
 | Aspect | Skills | Agents |
 |--------|--------|--------|
 | **Location** | `plugins/*/skills/name/SKILL.md` | `plugins/*/agents/name/AGENT.md` |
-| **Invocation** | Skill tool or slash commands | Task tool with `subagent_type` |
-| **Activation** | Automatic (based on keywords) | Explicit (you call them) |
-| **Required File** | `SKILL.md` with YAML frontmatter | `AGENT.md` or agent config |
-| **Purpose** | Expand context with knowledge | Execute multi-step tasks |
+| **Invocation** | `Skill()` or `/command` | `Task()` with `subagent_type` |
+| **Activation** | Automatic (keywords) | Explicit call |
+| **File** | `SKILL.md` (YAML frontmatter) | `AGENT.md` |
 
-**Correct Invocation Examples**:
+**Agent naming**: `{plugin}:{directory}:{yaml-name}`
 
 ```typescript
-// ✅ CORRECT: Invoking a skill
-Skill({ skill: "increment-quality-judge-v2" });
-// OR use slash command:
-/specweave:qa 0047
-
-// ✅ CORRECT: Invoking an agent
-Task({
-  subagent_type: "specweave:qa-lead:qa-lead",
-  prompt: "Create test plan for increment 0047"
-});
-
-// ❌ WRONG: Trying to invoke skill as agent
-Task({
-  subagent_type: "specweave:increment-quality-judge-v2",  // ERROR!
-  prompt: "Quality assessment"
-});
-```
-
-**Agent Naming Convention** (CRITICAL!):
-
-Claude Code agents follow a strict naming pattern for `subagent_type`:
-
-**Directory-based agents**: `{plugin-name}:{directory-name}:{name-from-yaml}`
-
-Examples:
-- Agent at `plugins/specweave/agents/qa-lead/AGENT.md` with `name: qa-lead`
-  → Invoke as: `specweave:qa-lead:qa-lead`
-- Agent at `plugins/specweave/agents/pm/AGENT.md` with `name: pm`
-  → Invoke as: `specweave:pm:pm`
-- Agent at `plugins/specweave/agents/architect/AGENT.md` with `name: architect`
-  → Invoke as: `specweave:architect:architect`
-
-**Why the "duplication"?**
-The pattern is `{plugin}:{directory}:{yaml-name}`. When the directory name matches the YAML `name` field (best practice), it creates the appearance of duplication: `qa-lead:qa-lead`.
-
-**File-based agents** (legacy pattern):
-- Agent at `plugins/specweave/agents/code-reviewer.md`
-  → Invoke as: `specweave:code-reviewer` (no duplication)
-
-**How to find the correct agent type**:
-```bash
-# List all available agents
-ls -la plugins/specweave/agents/
+// ✅ CORRECT: Skill
+Skill({ skill: "increment-quality-judge-v2" }); // or /specweave:qa
 
-# Check YAML name field
-head -5 plugins/specweave/agents/qa-lead/AGENT.md
-# Output: name: qa-lead
+// ✅ CORRECT: Agent
+Task({ subagent_type: "specweave:qa-lead:qa-lead", prompt: "..." });
 
-# Construct agent type: specweave:qa-lead:qa-lead
+// ❌ WRONG: Skill as agent
+Task({ subagent_type: "specweave:increment-quality-judge-v2" });  // ERROR!
 ```
 
-**Common Mistakes**:
-```typescript
-// ❌ WRONG: Missing the directory/name duplication
-Task({ subagent_type: "specweave:qa-lead", ... });
-// Error: Agent type 'specweave:qa-lead' not found
-
-// ✅ CORRECT: Full pattern with duplication
-Task({ subagent_type: "specweave:qa-lead:qa-lead", ... });
-```
-
-**Directory Structure Requirements**:
-
-```bash
-# ✅ CORRECT: Skill with SKILL.md
-plugins/specweave/skills/increment-quality-judge-v2/
-  └── SKILL.md  (with YAML frontmatter)
+**Validation**: `bash scripts/validate-plugin-directories.sh --fix`
 
-# ✅ CORRECT: Agent with AGENT.md
-plugins/specweave/agents/qa-lead/
-  └── AGENT.md
+**Incident**: 2025-11-20 - Empty agent directory caused "Agent not found" error
 
-# ❌ WRONG: Empty agent directory
-plugins/specweave/agents/increment-quality-judge-v2/
-  (empty - no AGENT.md!)
-
-# ❌ WRONG: Skill missing SKILL.md
-plugins/specweave/skills/my-skill/
-  └── some-file.txt  (not SKILL.md!)
-```
-
-**SKILL.md Format** (MANDATORY):
-```yaml
----
-name: skill-name
-description: What it does AND trigger keywords. Include all variations.
 ---
 
-# Skill Content
-
-Your skill documentation here...
-```
-
-**Validation** (MANDATORY before commits):
-
-```bash
-# Check for empty/invalid directories
-bash scripts/validate-plugin-directories.sh
-
-# Auto-fix empty directories
-bash scripts/validate-plugin-directories.sh --fix
-
-# This catches:
-# - Empty agent/skill directories
-# - Missing SKILL.md files
-# - Missing YAML frontmatter
-# - Invalid plugin structures
-```
+### 16. YAML Frontmatter Validation
 
-**Common Mistakes**:
-
-1. **Creating empty agent directories during scaffolding**
-   ```bash
-   # ❌ WRONG
-   mkdir -p plugins/specweave/agents/new-agent
-   git add . && git commit  # Empty directory!
-
-   # ✅ CORRECT
-   mkdir -p plugins/specweave/agents/new-agent
-   echo "---" > plugins/specweave/agents/new-agent/AGENT.md
-   # ... add agent content ...
-   git add . && git commit
-   ```
-
-2. **Copying skill as agent without changing structure**
-   - Skills need `SKILL.md` with YAML
-   - Agents need `AGENT.md` or config
-   - Don't copy-paste between them!
-
-3. **Using wrong invocation method**
-   - Check available skills: Skills list in context
-   - Check available agents: Error message shows all agents
-   - Skills → Skill tool or slash commands
-   - Agents → Task tool
-
-**When to Use Skills vs Agents**:
-
-**Use Skills when**:
-- Providing domain knowledge (e.g., Kafka best practices)
-- Expanding Claude's context with project-specific info
-- Explaining concepts, patterns, frameworks
-- Want automatic activation based on keywords
-
-**Use Agents when**:
-- Need multi-step task execution (e.g., generate docs)
-- Require specialized sub-agent capabilities
-- Building complex workflows with tools
-- Want explicit control over when they run
-
-**Incident History**:
-- **2025-11-20**: `specweave:increment-quality-judge-v2` agent invocation failed because only skill existed. Empty agent directory caused "Agent type not found" error. Fixed by removing empty directory and documenting distinction.
-
-**Prevention**:
-1. Run `bash scripts/validate-plugin-directories.sh` before commits
-2. Never create empty agent/skill directories
-3. Always include required files (SKILL.md, AGENT.md)
-4. Test invocation method matches component type
-5. Use pre-commit hook (blocks invalid structures)
-
-**See Also**:
-- Validation script: `scripts/validate-plugin-directories.sh`
-- Skills index: `plugins/specweave/skills/SKILLS-INDEX.md`
-- Claude Code Skills docs: `~/CLAUDE.md` (Personal skills reference)
-
-### 16. YAML Frontmatter Validation (CRITICAL!)
-
-**Rule**: ALL spec.md files MUST have valid YAML frontmatter with required fields.
-
-**Required Format**:
+**Required format**:
 ```yaml
 ---
-increment: 0001-feature-name  # REQUIRED: 4-digit number + kebab-case
-title: Feature Title           # OPTIONAL: Human-readable title
-feature_id: FS-001            # OPTIONAL: Living docs feature ID
+increment: 0001-feature-name  # REQUIRED: 4-digit + kebab-case
+title: Feature Title           # OPTIONAL
+feature_id: FS-001            # OPTIONAL
 ---
 ```
 
-**Common Mistakes**:
+**Common mistakes**: Unclosed brackets/quotes, invalid objects, missing `increment`, uppercase in ID
 
-```yaml
-# ❌ WRONG: Unclosed bracket
----
-increment: 0001-test
-data: [unclosed
----
-
-# ❌ WRONG: Unclosed quote
----
-increment: 0001-test
-title: "unclosed string
----
-
-# ❌ WRONG: Invalid YAML object
----
-increment: 0001-test
-config: {key: value, broken
----
-
-# ❌ WRONG: Invalid increment ID format
----
-increment: 1-test  # Missing leading zeros
----
-
-# ❌ WRONG: Invalid increment ID format (uppercase)
----
-increment: 0001-Test-Feature  # Uppercase letters not allowed
----
-
-# ❌ WRONG: Missing required field
----
-title: Feature Title  # Missing increment field!
----
-
-# ✅ CORRECT: Minimal valid frontmatter
----
-increment: 0001-feature-name
----
-
-# ✅ CORRECT: Full frontmatter
----
-increment: 0042-bug-fix
-title: Critical Bug Fix
-feature_id: FS-013
----
-```
-
-**Prevention Layers**:
-
-1. **Pre-commit hook** - Validates YAML before commits (blocks malformed frontmatter)
-   ```bash
-   # Runs automatically on git commit
-   scripts/pre-commit-yaml-validation.sh
-   ```
-
-2. **Spec parser** - Uses `js-yaml` library (detects errors at runtime)
-   - Provides descriptive error messages
-   - Shows line numbers for YAML errors
-   - Suggests common fixes
-
-3. **Command validation** - `/specweave:validate` checks YAML syntax
-   ```bash
-   /specweave:validate 0001
-   ```
-
-**Manual YAML Testing**:
+**Validation layers**:
+1. Pre-commit hook: `scripts/pre-commit-yaml-validation.sh`
+2. Spec parser (uses `js-yaml`, provides line numbers)
+3. `/specweave:validate 0001`
 
+**Manual test**:
 ```bash
-# Validate YAML in spec.md (manual check)
-node -e "
-  const yaml = require('js-yaml');
-  const fs = require('fs');
-  const content = fs.readFileSync('.specweave/increments/0001-test/spec.md', 'utf-8');
-  const frontmatter = content.match(/^---\n([\s\S]*?)\n---/);
-  if (!frontmatter) throw new Error('No frontmatter found');
-  const parsed = yaml.load(frontmatter[1]);
-  console.log('✅ Valid YAML:', JSON.stringify(parsed, null, 2));
-"
-```
-
-**Error Messages**:
-
-When YAML validation fails, you'll see descriptive errors:
-
-```
-ERROR: Malformed YAML syntax in frontmatter
-DETAILS: Unexpected end of stream
-LOCATION: Lines 2-5
-HINT: Common mistakes:
-  - Unclosed brackets: [unclosed
-  - Unclosed quotes: "unclosed
-  - Invalid YAML object: {key: value, broken
+node -e "const yaml = require('js-yaml'); const fs = require('fs'); \
+  const content = fs.readFileSync('.specweave/increments/0001-test/spec.md', 'utf-8'); \
+  const fm = content.match(/^---\n([\s\S]*?)\n---/); \
+  console.log('✅ Valid:', JSON.stringify(yaml.load(fm[1]), null, 2));"
 ```
 
-**Why This Matters**:
-- **Silent failures**: Malformed YAML can cause missing/incorrect metadata
-- **Sync issues**: Invalid increment IDs break living docs sync
-- **GitHub sync**: Broken frontmatter prevents issue creation
-- **Validation failures**: Tasks/ACs can't be parsed without valid metadata
-
-**Incident History**:
-- **Test case**: `tests/integration/commands/plan-command.integration.test.ts:626` tests malformed YAML handling
-- **Root cause**: Regex-based parsing was tolerant but unreliable
-- **Solution**: Multi-layered YAML validation with `js-yaml` library
-
-**See Also**:
-- Pre-commit hook: `scripts/pre-commit-yaml-validation.sh`
-- Spec parser: `src/generators/spec/spec-parser.ts` (uses js-yaml)
-- Test suite: `tests/integration/commands/plan-command.integration.test.ts`
-
 ---
 
 ## Project Structure
 
 ```
-src/                    # TypeScript code (compiled to dist/)
-plugins/                # Claude Code components (skills, agents, commands, hooks)
+src/                    # TypeScript (compiled to dist/)
+plugins/                # Skills, agents, commands, hooks
 ├── specweave/          # Core plugin
-└── specweave-*/        # Optional plugins (GitHub, JIRA, etc.)
-.specweave/             # Framework data (increments, docs, logs)
+└── specweave-*/        # Optional plugins
+.specweave/             # Increments, docs, logs
 ```
 
-**Rules**:
-- `src/` = TypeScript ONLY
-- ALL skills/agents/commands/hooks = `plugins/`
-- Marketplace = GLOBAL via CLI
-- NEVER mix `*.ts` and `SKILL.md` in same directory
-- NEVER create files in project root
+**Rules**: `src/` = TS only, ALL components = `plugins/`, NEVER mix `.ts` + `SKILL.md`, NEVER root files
 
 ---
 
-## Plugin Hook Registration (CRITICAL!)
-
-**When**: Adding or modifying hooks that respond to Claude Code tool events
-
-**Hook Schema Rules** (v0.22.14+):
-1. ✅ **ONLY use valid Claude Code hook events** (e.g., PostToolUse, PreToolUse, SessionStart)
-2. ❌ **NEVER use custom hook names** (e.g., "TodoWrite") as hook events
-3. ✅ **Use matchers** to filter which tool calls trigger the hook
-
-### Valid Claude Code Hook Events
+## Plugin Hook Registration
 
-Claude Code supports **only these 10 hook events**:
-- `PostToolUse` - After a tool completes (use this for TodoWrite, Write, Edit, etc.)
-- `PreToolUse` - Before a tool executes
-- `PermissionRequest` - When permission dialogs appear
-- `Notification` - When notifications are sent
-- `UserPromptSubmit` - When users submit prompts
-- `Stop` - When the main agent finishes
-- `SubagentStop` - When subagents finish
-- `PreCompact` - Before compaction operations
-- `SessionStart` - At session initialization
-- `SessionEnd` - When sessions terminate
-
-### Correct Hook Registration Format
-
-**File**: `plugins/*/. claude-plugin/plugin.json`
+**Valid hook events** (10 total): `PostToolUse`, `PreToolUse`, `PermissionRequest`, `Notification`, `UserPromptSubmit`, `Stop`, `SubagentStop`, `PreCompact`, `SessionStart`, `SessionEnd`
 
+**Format** (`plugins/*/.claude-plugin/plugin.json`):
 ```json
 {
-  "name": "specweave",
-  "version": "0.22.14",
   "hooks": {
     "PostToolUse": [
       {
         "matcher": "TodoWrite",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "${CLAUDE_PLUGIN_ROOT}/hooks/post-task-completion.sh",
-            "timeout": 10
-          }
-        ]
+        "hooks": [{
+          "type": "command",
+          "command": "${CLAUDE_PLUGIN_ROOT}/hooks/post-task-completion.sh",
+          "timeout": 10
+        }]
       }
     ]
   }
 }
 ```
 
-**Schema Breakdown**:
-| Field | Type | Purpose |
-|-------|------|---------|
-| `PostToolUse` | array | Hook event (one of 10 valid events above) |
-| `matcher` | string | Tool name pattern (e.g., "TodoWrite", "Write\|Edit") |
-| `type` | string | Always "command" for shell scripts |
-| `command` | string | Script path (use `${CLAUDE_PLUGIN_ROOT}` for plugin directory) |
-| `timeout` | number | Seconds before timeout (default: 30) |
-
-### Common Mistakes (v0.22.13 Bug)
-
-❌ **WRONG** (Invalid hook event):
-```json
-{
-  "hooks": {
-    "TodoWrite": {
-      "post": "./hooks/post-task-completion.sh"
-    }
-  }
-}
-```
-**Error**: `"TodoWrite"` is not a valid Claude Code hook event
-
-✅ **CORRECT** (Use PostToolUse with matcher):
-```json
-{
-  "hooks": {
-    "PostToolUse": [
-      {
-        "matcher": "TodoWrite",
-        "hooks": [{"type": "command", "command": "..."}]
-      }
-    ]
-  }
-}
-```
-
-### Testing Hook Registration
-
-```bash
-# 1. Update plugin.json with correct schema
-vim plugins/specweave/.claude-plugin/plugin.json
-
-# 2. Commit and push
-git add plugins/specweave/.claude-plugin/plugin.json
-git commit -m "fix: correct hook registration schema"
-git push origin develop
-
-# 3. Wait 5-10 seconds for marketplace auto-update
-
-# 4. Force refresh marketplace (if needed)
-cd ~/.claude/plugins/marketplaces
-rm -rf specweave
-git clone https://github.com/YOUR_USERNAME/specweave.git
-
-# 5. Restart Claude Code
-
-# 6. Verify no plugin loading errors
-# Should NOT see: "hooks: Invalid input" error
-```
-
-### Documentation References
-
-- **Claude Code Plugin Hooks**: https://code.claude.com/docs/en/hooks.md
-- **Hooks Reference**: https://code.claude.com/docs/en/hooks-reference.md
-- **Fix History**: See CHANGELOG.md v0.22.14 for the hook schema bug fix
+**❌ WRONG**: `"TodoWrite": {...}` (invalid event)
+**✅ CORRECT**: `"PostToolUse"` with `"matcher": "TodoWrite"`
 
 ---
 
-## Development Workflow
-
-### Core Commands
-
-```bash
-# Primary workflow
-/specweave:increment "feature"  # Plan new work
-/specweave:do                   # Execute tasks
-/specweave:progress             # Check status
-/specweave:done                 # Close (validates ACs, tasks, tests)
-
-# State management
-/specweave:pause 0002 --reason="..."
-/specweave:resume 0002
-/specweave:abandon 0002
-
-# Quality
-/specweave:validate 0001
-/specweave:qa 0001
-
-# Documentation
-/specweave:sync-docs update
-/specweave:sync-tasks
-```
-
-### Local Development Setup
-
-```bash
-# 1. Clone and install
-git clone https://github.com/YOUR_USERNAME/specweave.git
-cd specweave
-npm install
+## 9a. Hook Performance & Safety (CRITICAL - v0.24.3)
 
-# 2. Create symlink (CRITICAL!)
-mkdir -p ~/.claude/plugins/marketplaces
-rm -rf ~/.claude/plugins/marketplaces/specweave
-ln -s "$(pwd)" ~/.claude/plugins/marketplaces/specweave
+**Critical incidents**: 2025-11-22 - Multiple Claude Code crashes due to hook overhead
 
-# 3. Verify setup
-bash .specweave/increments/0043-spec-md-desync-fix/scripts/verify-dev-setup.sh
+**Root cause**: Process exhaustion from spawning 6+ Node.js processes per task completion
 
-# 4. Install git hooks
-bash scripts/install-git-hooks.sh
-```
+**Emergency fixes implemented (v0.24.3)**:
 
-**If "Plugin not found" errors**:
+### 1. Emergency Kill Switch
 ```bash
-echo '{"version": 1, "plugins": {}}' > ~/.claude/plugins/installed_plugins.json
-claude plugin marketplace update specweave
+# INSTANT disable of ALL hooks
+export SPECWEAVE_DISABLE_HOOKS=1
 ```
 
----
+### 2. Circuit Breaker (Auto-Protection)
+- **Threshold**: 3 consecutive failures → auto-disable hooks
+- **File**: `.specweave/state/.hook-circuit-breaker`
+- **Recovery**: `rm .specweave/state/.hook-circuit-breaker`
 
-## Build & Test
+### 3. File Locking (Prevents Concurrent Execution)
+- **Max instances**: 1 per hook type
+- **Timeout**: 5-10 seconds with stale lock cleanup
+- **Mechanism**: Directory-based mutex
 
-### Build
+### 4. Aggressive Debouncing
+- **Window**: 5 seconds (increased from 1s)
+- **Effect**: Batches rapid operations
+- **Trade-off**: 5s staleness acceptable for UX
 
+### 5. Complete Error Isolation
 ```bash
-npm run rebuild    # Clean + build (use this during development)
-npm run build      # Compile TypeScript + copy hook dependencies
-npm run clean      # Remove dist/
+set +e  # NEVER use set -e in hooks
+exit 0  # ALWAYS exit 0, never block workflow
 ```
 
-**Build Architecture**:
-1. `tsc` compiles `src/**/*.ts` → `dist/src/**/*.js`
-2. `copy:locales` copies translation files
-3. `copy:plugins` compiles plugin TypeScript with esbuild
-4. `copy:hook-deps` **NEW**: Copies hook dependencies to vendor/
+### 6. Consolidated Background Work
+- **Before**: 6+ Node.js spawns per task (exhaustion!)
+- **After**: 1 consolidated background job
+- **Reduction**: 85% fewer processes
 
-**Hook Dependencies** (NEW in v0.22.15):
-- Problem: Hooks imported from `../../../../dist/src/...` (failed in marketplace)
-- Solution: Copy compiled files to `plugins/*/lib/vendor/`
-- Hooks now import from `../vendor/...` (self-contained)
-- Script: `scripts/copy-hook-dependencies.js`
-- Auto-detect deps: `scripts/find-hook-dependencies.js`
+### Hook Safety Checklist (MANDATORY)
 
-**Critical**: Always import with `.js` extensions:
-```typescript
-// ✅ CORRECT
-import { foo } from './bar.js';
+**✅ EVERY hook MUST have**:
+1. Kill switch check (`SPECWEAVE_DISABLE_HOOKS`)
+2. Circuit breaker check (3 failure threshold)
+3. File locking (prevent concurrent runs)
+4. Debouncing (5s minimum)
+5. Error isolation (`set +e`, `exit 0`)
+6. Background work wrapped in subshell
+7. Circuit breaker updates on success/failure
 
-// ❌ WRONG
-import { foo } from './bar';
-```
+**❌ NEVER in hooks**:
+- `set -e` (causes crashes)
+- Synchronous Node.js spawns
+- Multiple separate background jobs
+- Error propagation to Claude Code
+- Missing `exit 0` at end
 
-**Fix missing extensions**: `node scripts/fix-js-extensions.js`
+### Performance Targets
 
-### Testing
+| Metric | Target | Warning | Critical |
+|--------|--------|---------|----------|
+| Hook execution | <100ms | 100-500ms | >500ms |
+| Background processes | 0-2 | 3-5 | 6+ |
+| Circuit breaker count | 0 | 1-2 | 3 (open) |
 
-**Framework**: Vitest (migrated from Jest 2025-11-17)
+### Emergency Recovery
 
+**If Claude Code crashes**:
 ```bash
-npm test                    # Smoke tests
-npm run test:unit           # Unit tests
-npm run test:integration    # Integration tests
-npm run test:e2e            # E2E tests (Playwright)
-npm run test:all            # All tests
-npm run test:coverage       # Coverage report
-```
-
-**Test structure**:
-- `tests/unit/` - Pure logic (no I/O)
-- `tests/integration/` - Organized by: `core/`, `external-tools/`, `generators/`, `features/`
-- `tests/e2e/` - Full user scenarios
-
-**Naming**: ALL tests use `.test.ts` (NEVER `.spec.ts`)
-
-### Writing Tests
-
-```typescript
-// ✅ CORRECT
-import { describe, it, expect, beforeEach, vi } from 'vitest';
-import fs from 'fs-extra';
-
-vi.mock('fs-extra');
-
-const mockReadFile = vi.mocked(fs.readFile);
+# 1. Immediate kill switch
+export SPECWEAVE_DISABLE_HOOKS=1
 
-beforeEach(() => {
-  vi.clearAllMocks();
-  mockReadFile.mockResolvedValue('content');
-});
-```
-
-**Critical anti-patterns**:
-```typescript
-// ❌ NEVER use jest APIs
-jest.fn()                    // Use: vi.fn()
-jest.mock()                  // Use: vi.mock()
-
-// ❌ NEVER use project root in tests
-const testRoot = path.join(process.cwd(), '.test-something');
-
-// ✅ ALWAYS use temp directory
-const testRoot = path.join(os.tmpdir(), 'test-' + Date.now());
-```
+# 2. Reset circuit breaker
+rm -f .specweave/state/.hook-circuit-breaker
 
-**Use test utilities**:
-```typescript
-import { createIsolatedTestDir } from '../test-utils/isolated-test-dir';
+# 3. Clear locks
+rm -rf .specweave/state/.hook-*.lock
 
-const { testDir, cleanup } = await createIsolatedTestDir('my-test');
-try {
-  // Test code here
-} finally {
-  await cleanup();  // ALWAYS cleanup
-}
+# 4. Rebuild
+npm run rebuild
 ```
 
-### Build Architecture
+**See**:
+- `.specweave/docs/internal/emergency-procedures/HOOK-CRASH-RECOVERY.md` (Complete recovery guide)
+- ADR-0060 (Three-tier optimization architecture)
+- `.specweave/increments/0050-*/reports/hook-crash-analysis.md` (Incident analysis)
 
-**Dual compilation**:
-- `tsc`: `src/` → `dist/src/` (source files)
-- `esbuild`: `plugins/**/lib/hooks/*.ts` → in-place `.js` (hooks only)
+---
 
-**Hook imports**:
-```typescript
-#!/usr/bin/env node
-import { ACStatusManager } from '../../../../dist/src/core/...';  // Use dist/
-```
+## Development Workflow
 
-**Increment script imports**:
-```typescript
-#!/usr/bin/env tsx
-import { ACStatusManager } from '../../../../src/core/...';  // Use src/
+**Core commands**:
+```bash
+/specweave:increment "feature"  # Plan
+/specweave:do                   # Execute
+/specweave:progress             # Status
+/specweave:done 0002            # Close (validates)
+/specweave:validate 0001        # Validate
+/specweave:qa 0001              # Quality check
+/specweave:pause/resume/abandon # State management
 ```
 
-**Running increment scripts**:
+**Local setup**:
 ```bash
-npx tsx .specweave/increments/####/scripts/script-name.ts
+git clone https://github.com/YOUR_USERNAME/specweave.git
+cd specweave && npm install
+mkdir -p ~/.claude/plugins/marketplaces
+ln -s "$(pwd)" ~/.claude/plugins/marketplaces/specweave
+bash scripts/install-git-hooks.sh
 ```
 
-### Coverage
-
-- Critical paths: 90%+
-- Overall: 80%+
-
 ---
 
-## Common Tasks
-
-### Adding Components
+## Build & Test
 
-All components go into `plugins/`:
-- Core: `plugins/specweave/{skills|agents|commands|hooks}/`
-- Plugins: `plugins/specweave-{name}/{skills|agents|commands}/`
-- Tests: `tests/integration/` or `tests/unit/`
+**Build**:
+```bash
+npm run rebuild    # Clean + build (development)
+npm run build      # Compile TS + copy deps
+```
 
-See `.github/CONTRIBUTING.md` for details.
+**Architecture**: `tsc` → `dist/src/`, esbuild → plugin hooks, copy deps → `plugins/*/lib/vendor/`
 
-### Updating Documentation
+**CRITICAL**: Always `.js` extensions (`import { foo } from './bar.js'`)
 
+**Test**:
 ```bash
-# Internal docs
-vim .specweave/docs/internal/architecture/hld-system.md
-
-# Public docs
-vim .specweave/docs/public/guides/user-guide.md
+npm test                    # Smoke
+npm run test:unit           # Unit
+npm run test:integration    # Integration
+npm run test:all            # All
+npm run test:coverage       # Coverage (80%+ required)
 ```
 
-### Configuration Management
+**Test rules**:
+- Use `vi.fn()` (NOT `jest.fn()`)
+- Use `os.tmpdir()` (NOT `process.cwd()`)
+- ALL tests = `.test.ts` (NEVER `.spec.ts`)
+- Use `createIsolatedTestDir()` helper
 
-**CRITICAL**: SpecWeave separates secrets from configuration (v0.24.0+)
-
-**Architecture**:
-- **Secrets** (tokens, passwords) → `.env` (gitignored, NEVER commit)
-- **Configuration** (domains, strategies, settings) → `.specweave/config.json` (committed to git)
+---
 
-**Why this matters**:
-- ✅ Team shares configuration via git
-- ✅ Secrets stay local (12-Factor App pattern)
-- ✅ Onboarding: `cp .env.example .env` (fill in your tokens)
-- ✅ Type-safe configuration with validation
+## Configuration Management (v0.24.0+)
 
-**Using ConfigManager**:
+**Secrets** (.env, gitignored) vs **Config** (.specweave/config.json, committed)
 
 ```typescript
 import { getConfigManager } from '../core/config/index.js';
 
-// Read configuration
 const configManager = getConfigManager(projectRoot);
 const config = await configManager.read();
+await configManager.update({ issueTracker: { provider: 'jira', domain: 'example.atlassian.net' }});
+```
 
-// Get specific value
-const domain = await configManager.get('issueTracker.domain');
-
-// Update configuration
-await configManager.update({
-  issueTracker: {
-    provider: 'jira',
-    domain: 'example.atlassian.net',
-    strategy: 'project-per-team'
-  }
-});
-
-// Write entire config (validates before writing)
-await configManager.write(newConfig);
+**What goes where**:
 
-// Set specific value
-await configManager.set('sync.enabled', true);
-```
+| Type | Location | Example | Committed? |
+|------|----------|---------|------------|
+| Tokens/Emails | `.env` | `JIRA_API_TOKEN=xyz` | ❌ |
+| Domains/Strategies | `config.json` | `"domain": "example.atlassian.net"` | ✅ |
 
-**Init flow separation** (Jira example):
+**Migration**: `node -e "require('./dist/src/cli/commands/migrate-config.js').migrateConfig({ dryRun: true })"`
 
-```typescript
-// Step 1: Save secrets to .env (gitignored)
-const secretEnvVars = getJiraEnvVars(credentials); // ONLY tokens/emails
-saveCredentialsToEnv(projectPath, 'jira', secretEnvVars);
+**See**: ADR-0050, `src/core/config/config-manager.ts`
 
-// Step 2: Save config to .specweave/config.json (committed)
-const configManager = getConfigManager(projectPath);
-const jiraConfig = getJiraConfig(credentials); // domain, strategy, projects
-await configManager.update(jiraConfig);
+---
 
-// Step 3: Generate .env.example for team onboarding
-generateEnvExample(projectPath, 'jira');
-```
+## Cache Management (v0.24.0+)
 
-**What goes where**:
+**Smart Caching with 24-Hour TTL**: Reduces API calls by 90% during init and sync operations.
 
-| Data Type | Location | Example | Committed? |
-|-----------|----------|---------|------------|
-| API Tokens | `.env` | `JIRA_API_TOKEN=xyz` | ❌ Never |
-| Emails | `.env` | `JIRA_EMAIL=user@example.com` | ❌ Never |
-| Domains | `config.json` | `"domain": "example.atlassian.net"` | ✅ Yes |
-| Strategies | `config.json` | `"strategy": "project-per-team"` | ✅ Yes |
-| Project Keys | `config.json` | `"projects": [{"key": "PROJ1"}]` | ✅ Yes |
-| Sync Settings | `config.json` | `"includeStatus": true` | ✅ Yes |
+### Cache Architecture
 
-**Validation**:
+**Location**: `.specweave/cache/`
+**TTL**: 24 hours (configurable)
+**Format**: JSON with timestamps
 
-ConfigManager validates configuration before writing:
+### Cached Data
 
-```typescript
-const result = configManager.validate(config);
-if (!result.valid) {
-  console.error('Validation errors:', result.errors);
-  // Example error:
-  // {
-  //   path: 'issueTracker.provider',
-  //   message: 'Invalid provider. Must be one of: none, jira, github, ado',
-  //   value: 'invalid-provider'
-  // }
-}
-```
+| Cache Key | Data | Use Case |
+|-----------|------|----------|
+| `jira-projects-{domain}` | Project list | JIRA init (auto-discovery) |
+| `ado-config` | Org/project/teams | ADO init (manual entry) |
+| `jira-{PROJECT}-deps` | Boards, components, versions | On-demand dependency loading |
+| `ado-{PROJECT}-deps` | Area paths, teams | ADO dependency loading |
 
-**Migration Tool** (for existing projects):
+### Cache Operations
 
-If you have an existing project using the old .env-only format, use the migration tool:
+**Automatic caching** (during init):
+- JIRA: `promptJiraCredentials()` caches selected projects
+- ADO: `promptAzureDevOpsCredentials()` caches org/project config
 
+**Manual cache management**:
 ```bash
-# Preview what will be migrated (dry run)
-node -e "require('./dist/src/cli/commands/migrate-config.js').migrateConfig({ dryRun: true })"
+# Refresh cache (bypass TTL)
+/specweave-jira:refresh-cache --all
+/specweave-ado:refresh-cache --all
 
-# Perform migration
-node -e "require('./dist/src/cli/commands/migrate-config.js').migrateConfig()"
+# Clean old caches
+/specweave:cleanup-cache --older-than 7d
 
-# Force migration even if not needed
-node -e "require('./dist/src/cli/commands/migrate-config.js').migrateConfig({ force: true })"
+# View cache statistics
+/specweave:cache-stats
 ```
 
-**What the migration tool does**:
-1. ✅ Analyzes `.env` file and classifies variables
-2. ✅ Creates backup: `.env.backup.TIMESTAMP`
-3. ✅ Updates `.env` (keeps only secrets)
-4. ✅ Creates/updates `.specweave/config.json` (adds config)
-5. ✅ Generates `.env.example` for team onboarding
+### Cache Security
 
-**Classification logic**:
-- **Secrets** (stay in .env): Variables containing `token`, `api_token`, `pat`, `secret`, `key`, `password`, `credential`, `auth`, or email addresses
-- **Config** (move to config.json): Everything else (domains, strategies, project keys, etc.)
+**Never cached**: API tokens, PATs, passwords (secrets stay in `.env`)
+**Always cached**: Non-sensitive config (domains, project keys, org names)
+**Atomic writes**: Temp file → rename pattern prevents corruption
+**Auto-recovery**: Corrupted cache auto-deleted, fallback to API
 
-**See also**:
-- ADR-0050: Secrets vs Configuration Separation
-- `src/core/config/config-manager.ts` - ConfigManager implementation
-- `src/core/config/config-migrator.ts` - Migration tool implementation
-- `tests/unit/core/config/config-migrator.test.ts` - Usage examples
+### Integration
+
+**CLI Helpers**:
+- `src/cli/helpers/issue-tracker/jira.ts`: JIRA project caching
+- `src/cli/helpers/issue-tracker/ado.ts`: ADO config caching
+
+**Core Module**:
+- `src/core/cache/cache-manager.ts`: TTL validation, atomic writes, corruption handling
+
+**Tests**:
+- `tests/integration/cli/helpers/cache-integration.test.ts`: 85%+ coverage
+
+**See**: ADR-0051 (Smart Caching with TTL), `src/core/cache/rate-limit-checker.ts`
 
 ---
 
 ## Troubleshooting
 
-**Skills not activating**: Check YAML frontmatter, restart Claude Code
-**Commands not working**: Verify plugin installed, restart Claude Code
-**Tests failing**: Run `npm run rebuild`, check test output
-**Root folder polluted**: Move files to `.specweave/increments/####/reports/`
-**Hooks failing**: Ensure changes are pushed to GitHub (Claude Code auto-updates marketplace every 5-10s). For symlink mode, see `.specweave/docs/internal/advanced/symlink-dev-mode.md`
+- **Skills not activating**: Check YAML frontmatter, restart Claude Code
+- **Commands not working**: Verify plugin installed, restart
+- **Tests failing**: `npm run rebuild`
+- **Root polluted**: Move to `.specweave/increments/####/reports/`
+- **Hooks failing**: Push to GitHub (auto-updates 5-10s)
 
 ---
 
 ## Quick Reference
 
-**Commands**:
-- `/specweave:increment "feature"` - Plan
-- `/specweave:do` - Execute
-- `/specweave:done 0002` - Close (validates)
-- `/specweave:status` - Show status
-- `/specweave:progress` - Check progress
-- `/specweave:validate 0002` - Validate
-- `/specweave:qa 0002` - Quality check
-- `/specweave:pause/resume/abandon` - State management
-- `/specweave:archive/restore` - Archiving (manual only)
-- `/specweave:sync-docs update` - Sync living docs
+**Commands**: `/specweave:increment`, `/specweave:do`, `/specweave:done`, `/specweave:progress`, `/specweave:validate`, `/specweave:qa`
 
-**Build**:
-- `npm run rebuild` - Clean + build
-- `npm test` - Smoke tests
-- `npm run test:all` - All tests
-- `npm run test:coverage` - Coverage
+**Build**: `npm run rebuild`, `npm test`, `npm run test:all`
 
-**File Structure**:
-- Source: `src/` (TypeScript), `plugins/` (skills/agents/commands/hooks)
-- Increments: `.specweave/increments/`
-- Docs: `.specweave/docs/internal/`, `.specweave/docs/public/`
-- Tests: `tests/` (unit, integration, E2E)
+**Structure**: `src/` (TS), `plugins/` (components), `.specweave/` (data), `tests/` (tests)
 
 **Remember**:
-1. Push changes to GitHub → Claude Code auto-updates marketplace (5-10s)
-2. Keep root clean (use increment folders)
-3. Test before committing
-4. Never delete .specweave/ directories
-5. Use `/specweave:done` (never manual metadata edits)
+1. Push → GitHub → Claude Code auto-updates (5-10s)
+2. Keep root clean
+3. Test before commit
+4. NEVER delete `.specweave/`
+5. Use `/specweave:done` (not manual edits)
+
+**See**: `.github/CONTRIBUTING.md`, https://spec-weave.com
 
-**See also**: `.github/CONTRIBUTING.md`, https://spec-weave.com