agents-md-generator

Installation
SKILL.md

AGENTS.md Generator

Create a small, project-specific instruction layer that agents can load on every task. Treat context as a limited budget: include durable rules that materially change agent behavior, and leave general engineering knowledge or detailed documentation out.

Run this skill in the main conversation. Do not use subagents unless the user explicitly approves the proposed agent count and scope after being warned that delegation can increase usage. Ask again before expanding an approved scope.

Core contract

  • Prefer AGENTS.md as the shared repository instruction file.
  • Generate only files the user requested. Ask before adding a compatibility file for another tool.
  • Keep the root file broadly applicable. Put subtree-specific rules in nested instruction files only when the target tool supports them.
  • Record verified facts, not assumptions or generic best practices.
  • Link to detailed project documentation with a task-specific read condition instead of copying it.
  • Never store secrets, credentials, private prompts, conversation transcripts, or the user's verbatim request.
  • Do not add tool configuration directories to .gitignore wholesale. Some contain intentionally versioned project configuration.
  • Do not modify global instruction files or user configuration while generating repository instructions.

Output budget

Installs
66
GitHub Stars
64
First Seen
Jan 24, 2026
agents-md-generator — thienanblog/awesome-ai-agent-skills