agent-output-audit

To verify AI-generated code meets task requirements, you need an independent audit that checks completed tasks against actual repository files, test results, and acceptance criteria instead of relying on self-reported success. This process confirms the implementation matches reality.

Independent AI task auditing is a verification process that checks AI-implemented work against task artifacts, baseline gate execution, and repository evidence. It works by performing command discovery, requirement-to-test mapping, and AI test-hygiene checks to return a verdict with supporting evidence.

To audit a pull request for flaky tests and AI test hygiene issues, you review baseline gates, perform flaky-triage handling, and check for test hygiene red flags. This examines the requirement-to-test mapping and repository evidence to validate the claimed success.

Yes, you can use an independent audit to verify Compozy loop task completion. The audit reviews task frontmatter, memory notes, and implementation evidence from completed AI runs to determine whether the reported status aligns with the actual repository artifacts.

AI code verification requires no external dependencies to run, but it needs access to task artifacts, test results, and repository evidence. It executes baseline gates and performs command discovery within your existing environment to validate the completed work.

You should not rely on AI self-reported task completion status when approving pull requests or closing Compozy loops without independent verification. If the task claims success but lacks clear requirement-to-test mapping or shows flaky test behavior, an evidence-based audit is necessary.