Skip to content

Latest commit

 

History

History
55 lines (42 loc) · 2.72 KB

File metadata and controls

55 lines (42 loc) · 2.72 KB

Definition of Done (DOD)

Gate: Must pass BEFORE deployment. An item that fails DOD cannot be deployed. The queue tail enforces this automatically.

Checklist

Code Quality

  • All new code has tests — every US has a corresponding test file
  • Tests pass — project's greenlight/test suite 100% green
  • No new test failures — pre-existing failures documented, zero regressions
  • Quality audit clean/sc:analyze --focus quality returns only Low findings (or none). All findings classified using FIPD taxonomy (Fix/Investigate/Plan/Decide) instead of severity-only. Investigate/Decide findings must include an Unknown: clause.
  • Low findings fixed/sc:cleanup --type all applied and verified

Architecture

  • Service-first — business logic in service layer, not route handlers
  • No constraint violations — checked against project's CLAUDE.md
  • Backward compatible — or all callers updated

External Writes

  • Dry-run before mutation — any skill that writes to files outside the current task scope (BACKLOG.md, ROADMAP.md, external systems) MUST preview changes before executing. Skills log intended writes to .claude-action before applying them. Autopilot: always dry-run external writes first, then execute.

Committed

  • Clean commit — descriptive message, atomic scope
  • No uncommitted changesgit status clean after commit
  • No secrets — no credentials, API keys, or tokens in committed code
  • Pre-commit hooks pass — all configured hooks green

Deployable

  • Application starts — server responds, no crash on boot
  • Existing functionality intact — no regressions

Verification

  • Verification pass/verification-before-completion run with evidence (test output, screenshots) before marking task done

Pipeline Housekeeping

  • Queue item checked[x] in TODO-Today.md
  • DONE-Today updated — item moved with timestamp
  • BACKLOG updated — source item marked completed or removed

Enforcement (Mandatory Queue Tail)

1. /sc:analyze "<changed files>" --focus quality    (findings use FIPD classification)
2. /production-code-audit       (M+ size tasks only — deep security/perf/arch)
3. /sc:cleanup --type all
4. /sc:test --coverage          (if new test files)
5. /commit-smart
6. deploy                       (project-specific)

FIPD reference: All finding outputs use Fix/Investigate/Plan/Decide classification. Definitions and uncertainty declaration rules: KNOWN_PATTERNS.md

If any step fails, autopilot pauses and reports the blocker.