a1024053774

a1024053774/design-integrity-guardrails

Skill Claude Code

Audit whether tests, evals, benchmarks, autoresearch tasks, generated artifacts, public APIs, or user-visible workflows provide independent behavioral evidence, including acceptance and release verdicts for model-driven Agent products. Use at a completion checkpoint, not for ordinary implementation or…

a1024053774/design-integrity-guardrails

Skill Claude Code

Review one frozen candidate change for structural shortcuts — swallowed or broad exception handling, fallback/retry paths, compatibility branches, parallel APIs, and patches that bypass root-cause ownership — or route to behavioral-acceptance-review when an evaluation surface changed. Use once at a completion…