verification-before-completion

To stop agents from claiming a bug fix is done without running tests, enforce evidence-based verification that requires running the full proof command and reading complete output before stating completion status.

Evidence-based verification in developer workflows is a gating mechanism that blocks completion claims by requiring fresh proof command output and exit status counts before a task can be declared passing or fixed.

To gate a commit with regression testing and lint build checks, run the full verification command, confirm complete output and failure counts, and only then express readiness status with explicit evidence.

Yes, you can use this for CI readiness checkpoints before creating a PR by requiring the full verification command to run and confirming its output proves the claim before declaring commit or PR readiness.

The best way to prevent premature completion claims during test gating is to block satisfaction language and success implications when verification is missing, partial, outdated, or inferred from previous runs.

Inferred verification causes false completion claims because it bypasses running the actual proof command, allowing agents to state work is fixed or passing based on assumptions rather than fresh output evidence.