| name | team-apply-guard |
| description | Use when implementing an approved OpenSpec change, progressing through tasks, updating task status, or continuing an incomplete implementation. |
Team Apply Guard
Core Principle
/team-apply is the coordination entry point. It dispatches to team-implementation-guard and TDD process.
When to Use
Use this skill when:
/team-apply <change-id> is executed- Continuing a previously incomplete implementation
- Progressing through tasks.md task by task
Do NOT use this skill when:
- Creating proposals or spec deltas (use
team-openspec-guard)
- Verifying completed work (use
team-verification-guard)
- Reviewing code before merge (use
team-review-guard)
- No OpenSpec change exists yet — create one with
/team-propose first
Process
- Accept change-id
- Read OpenSpec artifacts (proposal, design, tasks, specs)
- Use
team-implementation-guard to govern each task's implementation
- Follow TDD discipline
- Generate progress reports
Key Rules
- Task loop is autonomous — do NOT pause between tasks asking for permission
- Verify before marking complete
- Don't do anything out of scope
- STOP only for Risk Signals (see
team-implementation-guard)
Examples
Standard flow
User: /team-apply add-user-auth
AI:
1. Read openspec/changes/add-user-auth/tasks.md
2. Find 5 incomplete tasks
3. Pick task 1.1: Create User model
4. Write tests first: tests/models/test_user.py
5. Implement models/user.py
6. Run pytest: 8 passed
7. Update tasks.md: mark 1.1 complete
8. Continue to task 1.2...
Resume after interruption
User: /team-apply add-user-auth
AI:
1. Read tasks.md
2. Find tasks 2.1 and 2.2 complete, 3.1 incomplete
3. Resume from task 3.1
4. "Continuing add-user-auth, completed 4/5, current task 3.1"
Blocked by open questions
User: /team-apply add-payment
AI:
1. Read proposal.md
2. Find unresolved hard blocker:
- Q1: Payment gateway selection (Hard block)
3. Stop implementation: "Unresolved hard blocker exists. Please confirm payment gateway before continuing."
Relationship with Guard Skills
/team-apply
├── team-implementation-guard (implementation guardrails)
├── TDD guidance (test-first)
└── Progress reporting (done/in-progress/remaining/blocked)
Bad Cases
When this skill fails, record in badCases/ directory:
- Input: what the user said
- Wrong output: what the AI did that it shouldn't have
- Expected output: what it should have done
- New rule needed: how to prevent recurrence
See badCases/ for case history.
