receiptslisted

# receipts One question, asked before every completion claim: > **What did I run, and what did it print?** If there is no answer, the work is not done — it is *believed* to be done. Those are different things, and only one of them is safe to report. --- ## The rule **Never state that something works without an observation that shows it working.** Not "the change looks right". Not "this should fix it". An observation: a command that ran, an output that was read, a page that rendered, a request that returned. Three ways to satisfy it, in descending order of strength: 1. **Ran it** — the test suite, the build, the script, the actual failing repro 2. **Observed it** — the page in a browser, the log line, the API response, the file on disk after writing it 3. **Could not** — say so, plainly, in the same breath as the claim: *"Changed X. Couldn't run it here — the suite needs a database I don't have."* Option 3 is a perfectly good answer. Silently skipping to "done" is not. --- ## What counts as evidence, per claim | If you're about to say | You need | |---|---| | "tests pass" | The test command in this session, and its output read | | "N/N passing" | The run that printed those numbers. Not a count from memory, not from an earlier session | | "it builds" | The build command, and its exit/output | | "fixed" / "works now" | **The original failing case, re-run and now passing.** A build succeeding is not proof a bug is fixed | | "deployed" / "it's live" | The deploy