verification-before-completion

To prevent false completion claims before a pull request, enforce evidence-based verification by running the relevant test suites, linters, and builds, then inspecting exit codes and full outputs before reporting success.

Evidence-based verification is the process of running definitive commands like test suites or linters, reading their complete output and exit codes, and requiring concrete diffs or logs as proof before claiming a task is complete.

Ensure tests and linters pass before committing by executing the full verification commands, inspecting the complete output and exit codes, and rejecting hedged language or success reports without concrete evidence.

Yes, this verification approach works across development workflows including CI validations and regression checks by requiring fresh command execution, complete output inspection, and explicit evidence attachment before reporting any success.

Agent-only success reports lack observable proof and risk shipping regressions or incomplete fixes, so this approach rejects them along with hedged language like probably or should, demanding concrete output or VCS diffs instead.