/implement - Implementation Agent

Usage:
  /implement {ID}                    Implement a single task
  /implement auto {ID}               Implement with auto-chain enabled
  /implement -m {ID1} {ID2} ...      Implement multiple tasks in parallel
  /implement auto -m {ID1} {ID2}     Multi-task with auto-chain
  /implement -h, --help              Show this help message
  /implement -v, --version           Show version

Arguments:
  {ID}    Task ID (number) or task filename (e.g., 001-auth-jwt)

Options:
  auto    After implementation, automatically chain through:
          test → document → ship
  -m      Spawn parallel agents for each task

Examples:
  /implement 1                       # Implement task #1
  /implement 001-auth-jwt            # Using task filename
  /implement auto 1                  # With auto-chain
  /implement -m 1 2 3                # Three tasks in parallel

Next: /test {ID}

Workflow Skills v1.4.1
https://github.com/eljun/claude-skills

/implement auto {ID} → implement code
    ↓
/test (haiku)
    │
PASS → /document → /ship → PR + notify
FAIL → /implement (with test report) → retry

/implement -m 1 2 3
       │
       ├── Validate all tasks exist in TASKS.md
       ├── Check for file overlap warnings
       │
       └── Spawn in parallel (using Task tool):
           ├── Agent-1: /implement 1
           ├── Agent-2: /implement 2
           └── Agent-3: /implement 3
       │
       └── Wait for all agents to complete
       └── Report combined status

/implement auto -m 1 2 3

✓ Task 1: 001-auth-jwt.md (found)
✓ Task 2: 002-fix-portal.md (found)
✗ Task 3: Not found in TASKS.md → STOP if any missing

⚠️ Warning: Tasks 1 and 2 both modify src/auth/middleware.ts

Options:
1. Continue anyway (may need manual conflict resolution)
2. Run sequentially instead
3. Cancel

⚠️ Task 2 is blocked by Task 1
Recommendation: Run sequentially or resolve dependency first

═══════════════════════════════════════════════════════════════
MULTI-TASK PERMISSION CHECK
═══════════════════════════════════════════════════════════════

Tasks to implement in parallel:
  #1 {Task 1 name}
  #2 {Task 2 name}
  #3 {Task 3 name}

PERMISSIONS FOR BACKGROUND AGENTS
───────────────────────────────────────────────────────────────
✅ GRANTED (safe operations):
   • Read, Edit, Write files
   • Glob, Grep (search)
   • Git: add, commit, status, diff, checkout -b, log
   • npm/npx: install, run, build, test, lint

⚠️ REQUIRES APPROVAL (detected in tasks):
   • {e.g., "Database migration in Task #2"}
   • {e.g., "Schema changes in Task #1"}
   • (None detected)

🚫 BLOCKED (agents will stop if needed):
   • File deletion (rm, unlink, rimraf)
   • Destructive git (push --force, reset --hard, clean -f)
   • Database destruction (DROP, TRUNCATE)
   • System commands (sudo, chmod)

───────────────────────────────────────────────────────────────
If agents encounter blocked operations, they will STOP and
report back. You can then perform those manually.
═══════════════════════════════════════════════════════════════

Approve and start parallel implementation? (yes/no)

Task({
  subagent_type: "general-purpose",
  model: "opus",
  prompt: "/implement {ID}",
  run_in_background: true,
  allowed_tools: [
    "Read",
    "Edit",
    "Write",
    "Glob",
    "Grep",
    "Bash(git add *)",
    "Bash(git commit *)",
    "Bash(git status)",
    "Bash(git diff *)",
    "Bash(git checkout -b *)",
    "Bash(git log *)",
    "Bash(npm install *)",
    "Bash(npm run *)",
    "Bash(npx *)"
  ]
})

/implement [auto] [-m] {ID} [{ID2} ...]
       ↓
1. Parse arguments: detect "auto" flag, "-m/--multi" flag, task ID(s)
2. If multi-task → Run pre-flight checks, spawn parallel agents, exit
3. Resolve task ID → find docs/task/{ID}-{name}.md
4. Read task document
5. If "auto" flag → set Automation: auto in task doc
6. Check Automation field (manual | auto)
7. Move task to "## In Progress" in TASKS.md
8. ⚠️ MANDATORY: Invoke specialized skills (see Step 2 below)
   └── /vercel-react-best-practices (if installed + React code)
   └── /supabase-postgres-best-practices (if installed + DB code)
9. Implement following task document steps
10. Commit with [task-{ID}] prefix for traceability
11. Update status to "TESTING" when complete
       ↓
┌─── Automation Mode? ───┐
│                        │
▼ Manual                 ▼ Auto
Notify user              Invoke /test {ID}
Ready for /test

> **Automation:** auto

Input: /implement 1
TASKS.md row: | 1 | Dashboard Redesign | HIGH | [001-dashboard-redesign.md](...) |
Resolved: docs/task/001-dashboard-redesign.md

docs/task/{ID}-{task-name}.md

undefined

undefined

⛔ STOP: You detected a specialized skill is installed but did not invoke it.

This is a guardrail violation. You MUST:
1. Invoke the skill NOW before writing any code
2. Apply the best practices from the skill to your implementation
3. Only then proceed to Step 3

Skipping specialized skills leads to:
- Suboptimal code patterns
- Performance issues
- Security vulnerabilities
- Technical debt

undefined

undefined

[task-{ID}] {type}: {description}

{optional body}

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

undefined

undefined

// 1. Create the file
// 2. Add to index exports if needed
// 3. Import where used

// 1. Read the file first (always!)
// 2. Understand existing patterns
// 3. Make minimal, focused changes
// 4. Don't refactor unrelated code

undefined

undefined

> **Status:** TESTING
> **Completed:** {Date}
> **Implementation Notes:** {Any important notes for tester}

┌─────────────────────────────────────────────────────────────┐
│ ⚠️ SPECIALIZED SKILLS VERIFICATION                          │
├─────────────────────────────────────────────────────────────┤
│ □ Did task involve React/Next.js code?                      │
│   → If YES: Did you invoke /vercel-react-best-practices?    │
│                                                             │
│ □ Did task involve database/Supabase code?                  │
│   → If YES: Did you invoke /supabase-postgres-best-practices│
│                                                             │
│ □ If skill was available but NOT invoked:                   │
│   → STOP. Go back and invoke it. Apply best practices.      │
│   → Then return here to complete.                           │
└─────────────────────────────────────────────────────────────┘

Implementation complete for: #{ID} - {Task Title}

Files changed:
- path/to/file1.tsx (created)
- path/to/file2.tsx (modified)

Commits made:
- [task-{ID}] feat: Add authentication middleware
- [task-{ID}] feat: Add token refresh logic

Next Steps:
  /test {ID}              # e.g., /test 1
  /test {ID}-{task-name}  # e.g., /test 001-auth-jwt

Implementation complete for: #{ID} - {Task Title}

Files changed:
- path/to/file1.tsx (created)
- path/to/file2.tsx (modified)

[AUTO] Spawning /test with haiku model...

Multi-task implementation complete!

Results:
├── Task #1: ✓ Complete (3 commits)
├── Task #2: ✓ Complete (2 commits)
└── Task #3: ✗ Failed (blocked by missing API)

Next Steps (for successful tasks):
  /test 1                  # Test task #1
  /test 2                  # Test task #2
  /test 001-auth-jwt       # Using task name

Failed task requires attention:
  Task #3: See docs/task/003-feature-name.md for blocker details

┌────────────────────────────────────────────────────────────────┐
│ SPECIALIZED SKILLS ARE NOT OPTIONAL ONCE INSTALLED             │
├────────────────────────────────────────────────────────────────┤
│ 1. BEFORE writing code → Check if skills are available         │
│ 2. IF available AND relevant → INVOKE immediately              │
│ 3. APPLY best practices to your implementation                 │
│ 4. BEFORE completing → Verify you invoked required skills      │
│ 5. IF skipped → Go back, invoke, review code, fix issues       │
└────────────────────────────────────────────────────────────────┘