principle-prove-it-workslisted

# Prove It Works Verify every task output by checking the real thing directly. Do not infer from proxies, self-reports, or "it compiles." **Why:** Unverified work has unknown correctness. Indirect verification (file mtimes, output freshness, agent self-reports, cached screenshots) feels cheaper than direct observation. Acting on a wrong inference costs far more than checking the source. **The Iron Law: no completion claims without fresh verification evidence.** If the proving command was not run in this message, the claim cannot be made — and the rule covers paraphrases, satisfaction ("Done!", "Perfect!"), and implications of success, not just the exact words. **Pattern:** After completing any task, ask: "how do I prove this actually works?" The gate, before any status claim: identify the command that proves it → run it fresh and complete → read the full output and exit code → only then state the claim, with the evidence. If the output contradicts the claim, state the actual status instead. | Excuse | Reality | |--------|---------| | "Should work now" / "I'm confident" | Confidence ≠ evidence — run it | | "Linter passed" | Linter ≠ compiler ≠ tests | | "Agent said success" | Verify the artifact independently | | "Partial check is enough" | Partial proves nothing | | "Just this once" / "I'm tired" | No exceptions | For regression tests, verify red-green: run the new test (pass) → revert the fix (test MUST fail) → restore (pass). A test never seen failing proves nothing.