AI Code Verification

Verification discipline for AI-generated code: close the loop with tests/build/run before declaring done, detect hallucinated APIs and dependencies (slopsquatting), spot tautological tests, avoid premature abstraction, and run an adversarial review from a fresh perspective. Use when about to report any coding task as complete, when reviewing AI-written code, when adding new dependencies, or when the user says 本当に動く?, 検証して, AIが書いたコード. Applies proactively to the agent's own output. For general review standards use code-review-standard — this skill adds the AI-specific checks on top. Hand off diagnosing the cause of a failing test or build to systematic-debugging. Not for te…