verification-before-completion

To stop unverified bug fix claims, you enforce a mandatory verification gate that requires concrete evidence before any completion claim. This process identifies specific commands, executes them fully, reads all output, and verifies results against the claim to prevent unverified assertions.

Evidence-based verification works by enforcing a strict 5-step process that requires concrete command output and test results to support any assertion of completion. It identifies verification commands, executes them fully, reads all output, and matches the results against the initial claim.

A completion gate is a mandatory verification checkpoint that prevents unverified assertions by requiring concrete evidence and fresh output before any claim of task completion, bug fix, or test passing is accepted. It enforces honesty and trust by executing specific commands and reading all output.

No specific test frameworks are required as dependencies. The verification process works by identifying the specific commands needed for verification, executing them fully in your existing environment, and reading all output to verify the results against the completion claim.

An AI states a task is fixed without output because it lacks a mandatory verification process enforcing evidence-based claims. Implementing a strict 5-step completion gate requires executing specific verification commands and presenting the passing output rather than just stating the fix is complete.