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. **Pattern:** After completing any task, ask: "how do I prove this actually works?" Check the real thing, not a proxy: - Check process liveness directly, not indirectly through derived state - Read the actual value, not a cached or derived representation - When verification fails, suspect the observation method before suspecting the system Code and features: 1. Build it (necessary but not sufficient) 2. Run it and exercise the actual feature path 3. Check the full chain: does data flow from input to output? 4. For integrations, test the full communication path end-to-end Delegation: trust artifacts, not self-reports. When verifying delegated work, inspect the actual output artifact (git diff, file contents, runtime behavior), not the delegate's summary. Agents report what they intended, not always what happened. **Evidence goes to the human, in the human's frame.** Look at the thing the human would look at: the UI, the endpoint response, the terminal output, the log line, the released artifact. Then show it: a screenshot, an output paste, a log excerpt. Verification that never surfaces i