What problem does it solve?
Manually auditing a Claude Code project's harness structure for missing files, outdated rules, broken references, and complexity bloat is time-consuming and error-prone. This skill automates that diagnosis across 5 key layers, giving you a clear, prioritized list of issues to fix so your AI project setup stays lean and functional.
Core Features & Use Cases
- 5-Layer Structural Diagnosis: Checks for required file existence, complexity thresholds, drift between rules and actual usage, connection consistency for Forge Harness environments, and skill activity/pattern appropriateness.
- Optional Language Lint Mode: Scans skill and agent documentation for self-marketing language, cushion words, version brags, and missing "Done When" sections to improve documentation quality.
- Prescription Reports & Integrations: Generates prioritized M/S/R (Immediate/Strongly Recommended/Recommended) fix reports, integrates with weekly harvest-loop audits, and includes regression guards and PR change consistency checks to prevent harness modifications from breaking existing functionality.
- Use Case: If your Claude Code project's setup has grown bloated, rules are outdated, or skills aren't triggering as expected, run this skill to get an actionable list of exactly what to adjust first.
Quick Start
Request a harness structure diagnosis for your current Claude Code project to receive a prioritized list of structural issues and actionable improvement prescriptions.