codew-release-qa-sweep

Codewhale Release QA Sweep

No sweep, no "done". This skill owns nothing itself; it points at the two places that do.

• Gates and the release cut: follow docs/RELEASE_CHECKLIST.md from the canonical checkout on main (docs/RELEASE_RUNBOOK.md for the tools behind it). Native builds go through scripts/dev-cargo.sh / scripts/dev-test.sh (cw-gates). Hosted CI runs once, at the final release step, against the frozen candidate commit.
• Manual QA targets: cw-dogfood step 5. Unit and build gates do not cover the live TUI.

Regressions the manual QA must cover

Exercise each in a real terminal and record what you saw. The scenarios are described in cw-dogfood; the refs are the reports that motivated them.

• Six-worker fanout liveness, including Esc mid-fanout: #3216, #2211, and the Windows Terminal retest path #3289.
• Multi-terminal route isolation: #3227.
• Running-turn input contract: #3203.

Reporting

Each command, pass or fail, and the salient output line (test counts, the --version string, the check-versions.sh verdict). For manual QA, what you observed per target, citing the ref. Name what you skipped (for example, no display for TUI QA) rather than implying coverage you do not have.

Red flags / don't

• Don't claim "done", "passing" or "merge-ready" without command output.
• Don't skip the manual targets because the build is green.
• Don't tag, publish, create a GitHub Release, push artifacts or merge or close a PR or issue without approval. A green sweep is readiness evidence, not permission.
• Credit mechanics for contributor work: cw-land step 3.