What problem does it solve? Agent instructions like AGENTS.md, CLAUDE.md, and SKILL.md often fail to produce the intended behavior, but it is hard to tell whether the cause is wording, loading order, precedence, trigger conflicts, or missing evidence. This Skill diagnoses guidance systems against intended behavior and observed usage before any revision is made. ## Core Features & Use Cases - Behavior-first diagnosis: States the expected behavior and failure signals before inspecting artifacts, separating symptoms from root causes. - Cross-surface auditing: Examines instruction files, Skills, references, policy files, and client configuration together, including authority, precedence, triggers, and context cost. - Third-party trust analysis: Traces provenance, scripts, data access, and outbound destinations of imported Skills as one combined capability chain. - Use Case: When agents ignore a policy written in AGENTS.md, use this Skill to determine whether the file actually loaded, which instruction won precedence, and whether the fix should be a local edit or a structural reorganization. ## Quick Start Audit my AGENTS.md and SKILL.md files as one guidance system and explain why the agent is not following the retry policy, with evidence and a validation plan.
