Agent Done-Check Kit

Make your coding agent prove it's done. A free-to-start kit for Claude Code, Cursor and Codex: the Definition-of-Done rule (paste into CLAUDE.md or AGENTS.md), an always-on Cursor rule, a Claude Code Stop hook that runs your tests/lint/build when the agent says it's finished and sends it back to work if they fail, a GitHub Actions job that runs the same checks, and a log of the 9 test cases we ran against the script. Plain text and one Python file you can read in five minutes. Pay what you want, $0 is fine; if it saves you one fake "all tests pass", $5 is a nice thank-you.

Honest limits: it only enforces the commands you list, it can't verify deploys or emails, and the Stop hook is Claude Code only (the checks file and CI mode work with any agent). We tested the script on its own; try it once with a deliberately failing check before you rely on it.

What's inside: README, DEFINITION-OF-DONE.md, .cursor/rules/done-check.mdc, .claude/hooks/done_gate.py, .claude/done-checks.txt, .claude/settings.json, .github/workflows/done-gate.yml, VERIFY-LOG.txt. Refund: not applicable at $0; if you paid and it didn't help, reply and I'll refund.