health

🥷 The Health skill provides a comprehensive audit framework for evaluating AI agent configurations and the maintainability of AI-written code within a project. It is designed to detect misalignments in agent instructions, runtime tools, verifier surfaces, and overall coding practices, especially when agents ignore instructions, hooks or MCP fail, validation is missing, or code becomes difficult to maintain. By flagging issues according to severity, this skill helps teams maintain high standards for AI behavior, configuration consistency, and project sustainability without performing standard debugging or PR reviews.

The main features of this skill include tiered project assessment (Simple, Standard, Complex), durable context preflight checks, and a structured audit workflow that covers agent configurations, instruction surfaces, runtime tools, verifiers, and maintainability layers. It can read project-specific and global instructions, analyze memory patterns, and identify potential risks like code rot, unclear contexts, missing verification, or stale paths. Its budget-aware approach starts with a summary audit and escalates to a full deep audit only under specific triggers, preserving computational resources while ensuring thorough analysis when needed.

Use cases for the Health skill include auditing AI agent configurations for compliance with project standards, detecting potential AI coding issues, and ensuring maintainable AI-written code. It is ideal for developers, AI project managers, or teams using Codex, Claude, or similar AI agents who want to maintain high code quality, prevent configuration drift, and proactively address potential risks. This skill is particularly useful in projects with multiple contributors or complex CI/CD pipelines, where configuration consistency and maintainability are critical.