verification-before-completion

Evidence-based reporting requires running fresh verification commands, inspecting their outputs, and confirming results before declaring tasks complete. This prevents relying on assumptions, partial checks, previous runs, or delegated agent-reported success.

You must verify code with fresh evidence before reporting a bug fix, committing code, creating pull requests, or declaring requirements complete to ensure status claims are backed by actual test and build outputs.

The best way to ensure pull requests are backed by verified test results is to execute relevant verification commands and inspect their actual outputs immediately before completion, avoiding reliance on partial checks or previous runs.

Yes, verification before completion works for delegated agent work and automated builds by requiring execution of relevant verification commands, inspection of outputs, and evidence-backed status reporting before any completion statements are made.

Relying on previous test runs causes inaccurate software quality assurance because it skips fresh verification, allowing false completion claims based on outdated or partial checks rather than current, evidence-backed command outputs.