fleet-completion-checklist
Pre-completion verification — everything to check before calling fleet_task_complete. Prevents incomplete submissions.
SKILL.md.claude/skills/fleet-completion-checklist/SKILL.md
GitHubFleet Completion Checklist
Before You Call fleet_task_complete
Go through this checklist. Every item must be verified — not assumed. If an item fails, fix it before completing.
1. Verbatim Requirement Check
- Read your task's
requirement_verbatimfield - For EACH acceptance criterion in the verbatim: can you point to specific code/commit that addresses it?
- If verbatim is empty: STOP — this is a problem. Post a comment asking PM to set it.
2. Plan Adherence
- Your commits match the plan you accepted via
fleet_task_accept - You did NOT add scope beyond the plan (no "while I'm here" changes)
- You did NOT modify files outside the plan's target files (unless dependencies required it)
3. Contributions Consumed
- Check: did you receive all required contributions? (use
eng_contribution_check) - Architect's design_input → your implementation follows the specified patterns
- QA's test_definition → your implementation passes all TC-XXX criteria
- DevSecOps' security_requirement → your implementation satisfies all controls
- If any required contribution is missing: do NOT complete. Call
fleet_request_input.
4. Tests
- Tests exist for new functionality
- All tests pass (run them — don't assume)
- Run:
.venv/bin/python -m pytest fleet/tests/ -v --tb=short - If tests fail: fix them. If you can't, create a blocker task.
5. Code Quality
- Conventional commits:
type(scope): description [task:XXXXXXXX] - Type hints on public functions
- No hardcoded paths or secrets
- No TODO/FIXME left without a follow-up task created
6. Delivery Phase Standards
Check your task's delivery_phase against config/phases.yaml:
| Phase | Tests | Docs | Security |
|---|---|---|---|
| poc | Happy path | README | No secrets in code |
| mvp | Main flows + edges | Setup + usage | Auth + validation |
| staging | Comprehensive | Full docs | Dep audit, pen-test mindset |
| production | Complete | Everything | Certified, compliance-verified |
If your work doesn't meet the phase standard: either improve it or flag to PM that the phase may need reassessment.
7. Summary Quality
Your fleet_task_complete(summary=...) summary should answer:
- What was done (specific files, specific changes)
- Why it was done (references the verbatim requirement)
- How it can be verified (specific test command or steps)
Bad: "Implemented the feature" Good: "Added JWT auth middleware in fleet/core/auth.py with RS256 signing per architect's design input. 5 tests in test_auth.py cover valid/invalid/expired tokens. Run: pytest fleet/tests/core/test_auth.py"
8. What Happens When You Call fleet_task_complete
The tool fires a 12+ operation tree:
- Push branch to remote
- Create PR (publication-quality via fleet-pr skill)
- Update task status → review
- Create approval for fleet-ops
- Post completion comment on task
- Notify IRC #reviews and #fleet
- Send ntfy notification to PO
- Update Plane issue state + labels
- Record trail event
- Notify contributors (QA, DevSecOps) their input is in review
- Evaluate parent task (all children done → parent to review)
- Update sprint progress
You do NOT need to do any of these manually. The tree handles it. But fleet-ops WILL review your work against this checklist. Submitting incomplete work wastes a review cycle and gets rejected.
Quick Decision Tree
Did I read the verbatim requirement?
├── No → read it first
└── Yes → Does every criterion have code addressing it?
├── No → implement the missing parts
└── Yes → Did I consume all required contributions?
├── No → fleet_request_input for missing ones
└── Yes → Do all tests pass?
├── No → fix the tests
└── Yes → Does the summary explain what/why/how?
├── No → write a proper summary
└── Yes → fleet_task_complete(summary)