verification-gate

Task completion verification requires running fresh test, build, lint, or type check commands and inspecting their exit codes and output to support your done claim. This process blocks false completion assertions by enforcing evidence-backed verification across multi-language codebases.

Anti-hallucination verification blocks low-evidence completion phrases by requiring real command output as proof. It enforces a five-step workflow: identify proof, run fresh verification commands, inspect full output, confirm it matches the claim, and provide an evidence-backed assertion.

Yes, verification supports multi-language codebases through language-aware command selection that auto-detects project types including Python, Node/TypeScript, Rust, and Go. It runs the appropriate test, build, lint, or type check commands for the detected project directory.

Use rollback guidance to reset the working tree and clear caches when verification commands generate unwanted changes. This keeps your project directory clean while enforcing fresh execution of test, build, lint, or type check commands for completion verification.

You need evidence-backed verification whenever completing tasks like bug fixes, new features, refactors, build fixes, lint fixes, or type fixes. It enforces fresh execution of appropriate verification commands with strict exit-code and output confirmation before allowing any completion claim.