← ClaudeAtlas

ccqa-resolvelisted

Clear the ccqa rows that are waiting on a person: read the hub's verdict for a project, drop what is already in hand, and route each remaining row to the repair it actually needs — a product fix, a re-recording, an environment repair — heaviest first. Use when asked to triage, resolve or work through ccqa findings, a release verdict, or "what needs attention".
shibukazu/ccqa · ★ 4 · AI & Automation · score 68
Install: claude install-skill shibukazu/ccqa
# Resolve what ccqa is holding for a person You are asked to clear what a project's ccqa hub says a person has to handle. Finish with every row either repaired or stopped at a stated reason. **The hub's verdict is the input.** If you were handed a notification, use it to find which project and profile to ask about — then work from what the hub answers now, not from the message. ## Rules that hold throughout - **Never weaken a test to clear a row.** Loosening an assertion or dropping a step turns a row green by destroying what made it worth reading. - **Never edit a generated artefact** — `test.spec.ts`, `ir.json`, `generated.json`, anything a target emits. A test that no longer fits the product is repaired by re-recording it, never by hand. - **Attesting and dismissing are a person's word, and each answers one thing.** `ccqa hub attest <feature>/<spec> --profile <profile> --by <name>` answers an `ENVIRONMENT` failure: somebody fixed the environment and checked the behaviour by hand. `ccqa hub dismiss <feature>/<spec> --by <name> --reason <text>` answers an audit finding that is wrong. Nothing answers a `TEST_DRIFT` or `PRODUCT_BUG` row — those are repaired, not overruled, and the CLI will not stop you getting that wrong. Hand a person the command with your reasoning; do not run it yourself, and never offer an attestation for a row your own run cleared, which would put a person's word behind a machine result. - **A run reaches the ledger only if it