Implementation Integrity Audit

Context

This skill is an adversarial quality audit for implementation honesty. It looks for bugs that are easy to miss when the diff appears complete: fake green paths, mocks with production names, handler failures logged as success, tests that assert fixtures rather than behavior, swallowed errors, and runtime evidence that does not prove the claim being made.

Use it when a previous review missed obvious issues, when agent-generated code may have optimized for passing tests, or when mocks/fakes/simulated runtimes are present near a production-facing boundary.

Inputs

• intake-brief: Target context, evidence requirements, non-targets, and safety constraints.
• source-code: The implementation diff and related runtime boundaries.
• test-suite: Unit, integration, E2E, fixture, smoke, and benchmark evidence.
• task-list: The exact work items and priority sequence.

Also inspect logs, debug records, runbooks, mocks, fixture data, generated artifacts, and quality reports when available.

Process