verification-before-completion

Run the full verification command for your claim, such as the test suite or build command, and read the complete output including exit code. Only state the change works after the output confirms it with zero failures.

Evidence is the fresh output of the test command showing zero failures from the current session. Previous runs, partial checks, or assumptions that code should pass do not qualify as verification.

Linters check code style and syntax patterns but do not compile or bundle the project. A build claim requires running the actual build command and confirming it exits with code 0.

Use the red-green cycle: write the test, confirm it passes with the fix, revert the fix and confirm the test fails, then restore the fix and confirm it passes again. A test that never fails proves nothing.

No, agent success reports must be independently verified by checking the version control diff and confirming the actual changes exist. Trusting reports without inspection is listed as a common failure mode.

Apply it before any success claim, expression of satisfaction, commit, pull request, task completion, or handoff to another agent. The rule covers exact phrases, paraphrases, and any implication of completion.