maintain-agents-md

Creates, reviews, and updates AGENTS.md and CLAUDE.md project-instructions files.

Updated Apr 17, 2026
One-click install
npx skills add https://github.com/hpark0011/mirror --skill maintain-agents-md-hpark0011
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: maintain-agents-md
Source: https://github.com/hpark0011/mirror/tree/main/.agents/skills/maintain-agents-md
Command: npx skills add https://github.com/hpark0011/mirror --skill maintain-agents-md-hpark0011

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Project-instructions files like AGENTS.md and CLAUDE.md often become bloated, stale, or duplicated across tools, wasting context tokens and reducing how reliably AI coding agents follow them. This Skill keeps those files terse, accurate, and aligned with official Claude Code memory conventions. ## Core Features & Use Cases - Create: Analyze a codebase (tech stack, commands, structure, conventions, gotchas) and generate a new CLAUDE.md under 200 lines with only verified commands. - Review: Audit an existing file and return a scored report with line-referenced issues, severity levels, and concrete rewrites. - Update: Apply targeted, diff-previewed changes to an existing file while preserving well-written content. - Use Case: Your repo already has an AGENTS.md for other agents and you want Claude Code to share it — this Skill makes CLAUDE.md a one-line @AGENTS.md import plus Claude-specific overrides, avoiding drift between the two files. ## Quick Start Ask the agent to review my AGENTS.md file and suggest improvements for context efficiency.

Frequently Asked Questions about maintain-agents-md

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a CLAUDE.md file for my project?▼

Run the create workflow, which scans your repo for config files like package.json or pyproject.toml, extracts verified build, test, and lint commands, maps key directories, and generates a terse CLAUDE.md under 200 lines following the official template.

How do I review or audit an existing CLAUDE.md?▼

The review workflow loads your CLAUDE.md alongside package.json and the directory structure, counts instructions, checks anti-patterns like verbosity and staleness, verifies every command exists, and returns a scored report with line-referenced fixes.

Should I use AGENTS.md or CLAUDE.md for project instructions?▼

Claude Code reads CLAUDE.md, not AGENTS.md. If your repo already uses AGENTS.md for other agents, make CLAUDE.md a one-line @AGENTS.md import plus any Claude-specific overrides so both tools share a single source of truth.

What is the recommended length for a CLAUDE.md file?▼

Keep CLAUDE.md under 200 lines, since it loads in full at every session start. Files between 200-300 lines should be split via @imports or .claude/rules/, and files over 300 lines are considered critical.

When should I use .claude/rules/ instead of CLAUDE.md?▼

Use .claude/rules/ for modular or rarely-triggered instructions. Rules with a paths: frontmatter are path-scoped and only load when Claude reads matching files, keeping edge-case rules out of the always-loaded context.