verification-before-completion

To prevent false completion claims in agent workflows, require fresh verification evidence like test runs and build outputs before updating any task status to done. You must run the full validating command, read complete outputs and exit codes, and confirm the evidence before declaring success.

Evidence-first verification for bug fixes is the practice of running the full validating command and reading its complete output and exit codes to confirm a fix actually works, rather than trusting status reports or assumptions before declaring the bug resolved.

To verify delegated agent work before committing, check the resulting diff and validation results instead of trusting status reports. Run the relevant proving command, read the fresh output, and confirm the evidence proves the work is complete before declaring success.

Yes, verification before completion works for regression checks and requirement reviews by applying the same discipline. It requires running the full validating command, reading complete output and exit codes, and confirming the evidence before declaring any status update across software delivery workflows.

You need to read complete output and exit codes for quality assurance because relying on partial output or assumptions leads to false completion claims. Fresh verification evidence ensures that tests, builds, and fixes are actually proven before any success status is declared.