Harness Health Audit

What it does

This skill performs a structured audit of a Claude Code harness, scoring eight distinct dimensions (CLAUDE.md quality, skills coverage, agents architecture, hooks automation, rules structure, MCP server capabilities, evaluation pipelines, and multi-agent/team setup). It produces a numerical score (0–24) and a graded rating (S/A/B/C/D) with actionable improvement recommendations.

When to use it

Use this skill when onboarding or reviewing an agent harness, before production deployments, or during regular maintenance to identify gaps in architecture, automation, or testing. It is triggered by requests related to harness audits, configuration checks, or health diagnostics.

What's included

• Scripts: none detected in the skill directory
• References: none provided, but the skill defines a clear scoring rubric and output format for programmatic reporting
• Instructions: Procedural scoring for eight dimensions with output formatted as a summary table and prioritized remediation suggestions.

Compatible agents

Designed for Claude Code harnesses and similar agent frameworks that expose CLAUDE.md, skills, and configuration files for inspection. It assumes the auditing agent can read local harness files and compute scores programmatically.