verification-before-completion

To prevent false claims of task completion, enforce a strict verification protocol that mandates running specific verification commands and reviewing their output to gather evidence before making any success assertions.

The evidence before assertions principle requires running relevant test suites and confirming the actual output before claiming code is fixed or passing, ensuring accuracy and preventing regressions in software development workflows.

To verify tests pass before committing, run the relevant verification commands, review their output for evidence, and only communicate completion assertions after confirming the test suite and linters are clean.

Yes, agentic task completion requires mandatory verification to prevent dishonesty, ensuring verification commands are executed and their output is reviewed as evidence before any success status is reported.

Assuming tests pass without verification leads to incomplete fixes because it skips gathering required evidence from verification command output, allowing regressions and false success claims to be introduced into the codebase.