verification-before-completion

Run verification commands for tests, linters, builds, and regression checks fully, then review outputs and exit codes before opening a pull request. This ensures completion claims are backed by fresh evidence rather than assumptions or partial checks.

Evidence-based reporting distinguishes verified results from assumptions or agent reports by requiring fresh command output and exit status before claiming work is complete, fixed, or passing. It prevents unsupported correctness or readiness claims in development workflows.

Require a verification gate that identifies and runs complete testing commands, reviews outputs and exit codes, and reports gaps when evidence is insufficient. This prevents delegated agent work from reporting unverified success without fresh test execution.

Yes, you can apply this verification approach to build validation and regression testing by running the relevant commands fully, checking diffs against requirements, and reporting only outcomes supported by fresh command output and exit status.

The best way to validate requirements is to identify the specific command proving the fix, run it fully, review its output and exit status, and check diffs against requirements. Report only outcomes supported by this fresh evidence rather than confidence-based conclusions.

Code review alone cannot substitute for build validation; you must run complete verification commands, review their outputs and exit codes, and check diffs against requirements to identify gaps where evidence is insufficient for completion claims.