Reviews, diagnoses, and creates AGENTS.md and CLAUDE.md agent instruction files.
npx skills add https://github.com/chujianyun/skills --skill agent-md-advisor
Please help me install this Agent Skill. Skill: agent-md-advisor Source: https://github.com/chujianyun/skills/tree/main/skills/review/agent-md-advisor Command: npx skills add https://github.com/chujianyun/skills --skill agent-md-advisor
AI coding agents often receive bloated, vague, or misconfigured instruction files like AGENTS.md and CLAUDE.md, which reduces their reliability. This Skill answers best-practice questions, diagnoses existing files against a structured rubric, and rewrites or creates concise, project-specific instruction files.
Ask the agent to review your AGENTS.md and CLAUDE.md against best practices and suggest focused improvements.
High-intent search queries and answers about installing and using this skill.
A good CLAUDE.md is a short technical brief of 80-120 lines covering project context, exact commands, architecture map, critical rules, and safety boundaries. Avoid persona prompts, generic advice, and pasted documentation; every line should prevent a concrete agent mistake.
AGENTS.md is a tool-neutral standard for cross-tool AI coding instructions, while CLAUDE.md is Claude Code project memory. For multiple tools, use AGENTS.md as the source of truth and add CLAUDE.md or a symlink only when tooling requires it.
AGENTS.md should stay under 150 lines, and CLAUDE.md ideally 80-120 lines with 200 as a soft ceiling. Files over 300 lines should be split into referenced docs, nested files, or rules using progressive disclosure.
No. Keep personal habits and machine-specific settings in CLAUDE.local.md or user-level memory. Shared CLAUDE.md files should contain only stable project facts like commands, architecture, and team-wide constraints.
Long files with generic or low-signal content reduce instruction-following reliability. Trim obvious rules, reserve MUST/MUST NOT for under 15 critical constraints, and verify that commands match real project scripts.
bmad-code-orgCreates and maintains verified AGENTS.md instruction blocks for AI coding agents in a repository.
hb9397Audits project documentation against source code for dependency, pattern, and rule discrepancies.
kid-sidConfigures Claude Code environments with skills, agents, hooks, permissions, and MCP servers.
Habitat-ThinkingSyncs HARNESS.md conventions to Cursor, Copilot and Windsurf rule files.
