Gate “done” claims behind a real test run

Goal: Stop trusting an agent’s word that the work is finished.
Setup: agents used in source: Claude, Codex, Cursor, Gemini CLI; tools/MCPs stated: Stop hook.
Steps:
1. Measure first: scan your local transcripts for “done” / “tests pass” claims and check them.
2. Add a Stop hook (Claude Code, Codex CLI and others have one) that runs checks when the agent claims completion: fast checks always, the full suite only when the claim warrants it.
3. Watch for green-without-fix: skipped or deleted tests, weakened assertions.
4. In CLAUDE.md / AGENTS.md: “Before claiming checks pass, re-run them after your last edit” — ask for proof, not a verdict.
Rules: ask me before anything irreversible (sending, posting, paying, deleting); don’t invent data, results or tools I haven’t given you; tell me what you couldn’t verify.
Based on: Raimondas Lencevicius — https://dev.to/raimondasl/69-of-my-coding-agents-done-claims-werent-here-is-the-gate-i-put-in-front-of-them-lho

Steps are from the source. Includes a “Based on” credit line.