The agents-md skill helps developers create and maintain AGENTS.md (and symlinked CLAUDE.md) files that follow research-backed best practices for minimalist documentation. By enforcing a strict line limit and focusing on high-signal information like file-scoped commands, package managers, and commit attribution, this skill prevents instruction-following degradation and token bloat. It intelligently analyzes your project's toolchain and repository layout to provide the most relevant context for Claude Code and other agentic IDEs without duplicating existing linter or formatter configurations.
主な機能
01Intelligent project analysis of lockfiles, linters, and CI configs
0240,350 GitHub stars
03Automated setup and symlinking of AGENTS.md and CLAUDE.md
04Standardized AI commit attribution templates for team transparency
05Enforcement of high-signal, minimal documentation (under 100 lines)
06Generation of optimized, file-scoped commands for faster agent execution
ユースケース
01Standardizing developer environment commands across large monorepos
02Initializing a repository for peak performance with Claude Code or Cursor
03Pruning bloated project instructions to improve agent reliability