← ClaudeAtlas

final-checklisted

Run the final AI-SDD self-review before handing off backend work — run format, static analysis, and targeted tests, verify the Definition of Done, and produce the handoff summary.
YasMax91/groundwork · ★ 0 · Code & Development · score 66
Install: claude install-skill YasMax91/groundwork
# Final check Run the gates and the self-review before declaring the work done. Use the project's commands (from `.groundwork.json`, defaulting to Sail). ## Run the gates - Format: `./vendor/bin/sail composer format:test` (or `format` to apply). - Static analysis: `./vendor/bin/sail composer analyse`. - Tests: the narrowest relevant first, then broader — e.g. `./vendor/bin/sail artisan test --filter=...`, then `./vendor/bin/sail composer test`. They must be **green** — the Stop gate also runs the suite on changed PHP (`gates.test_on_stop`). Confirm the tests for the changed behavior were written **test-first** (red→green). - OpenAPI: if the change touched an endpoint, regenerate the spec (`./vendor/bin/sail artisan l5-swagger:generate`) — it must finish with **no errors and no warnings**, and every touched endpoint must be documented to the standard in `${CLAUDE_SKILL_DIR}/../../guidelines/openapi-protocol.md`. The `openapi` Stop gate enforces this; if the change provably does not touch the contract, record `- OpenAPI: n/a — <reason>` in the checkpoint rather than working around the gate. Report results honestly. If a gate could not run (Sail/services unavailable), say so explicitly. ## Live verification (green gates are necessary, not sufficient) The gates prove the code satisfies the tests and the static rules; they do **not** prove the running app does what the user will see. Before "done", exercise the changed behavior against the **running** app the