{
  "id": 10912461,
  "title": "Awesome AGENTS.md That Can Be Used on Any Agent and Any Stack",
  "url": "https://urgent.news/2026/09/30/awesome-agents-md-that-can-be-used-on-any-agent-and-any-stack",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-09-30T00:21:07.000Z",
  "source": {
    "name": "HackerNoon",
    "slug": "hackernoon",
    "url": "https://hackernoon.com/awesome-agentsmd-that-can-be-used-on-any-agent-and-any-stack?source=rss"
  },
  "original_language": "en",
  "account": "The author reverse-engineered well-known coding-agent instruction repositories and created a stack-independent AGENTS.md file. This file contains core rules covering aspects such as small focused changes, security boundaries, debugging, and verification before an agent determines a task is complete. Additional specialized guidance resides in the rules/ directory, meant to be consulted when a task necessitates it. The core rules apply universally to any agent regardless of the stack or framework, enabling consistent behavior across Claude Code, Codex, Gemini, and Cursor. The core emphasizes understanding the request, maintaining the scope of changes, handling sensitive data, debugging failures, and verifying the result, ensuring actions violating boundaries like deleting files, rewriting Git history, or forcing a failing check pass are prevented without explicit permission. The agent is instructed to avoid rewriting code or passing failing checks by weakening them. Before incorporating a helper or dependency, the agent is advised to check if the solution can be simpler using existing code in the project, standard libraries, or dependencies. Verification involves running relevant checks, examining the output, and reporting the result before concluding the task. A module system in the rules/ directory offers framework-agnostic specialized rules for various tasks like database management, backend security, caching, deployment, payments, and browser automation. Each module explicitly states its scope and applies general principles where relevant. The core AGENTS.md file is concise, under 200 instruction lines and 32 KiB, ensuring it remains manageable. CI checks verify the file adheres to these size limits. The author encourages usage by cloning the repository, loading the core instructions for Claude Code as a plugin, or copying the core into Cursor's User Rules. For Gemini CLI, the core is added to GEMINI.md, and for OpenAI Codex CLI, the core is included in AGENTS.md. To confirm successful loading, the agent should conclude its responses with \"✓ awesome-agents-md.\"",
  "summary": "I reverse-engineered popular coding-agent instruction repos and distilled their best ideas into a stack-agnostic AGENTS.md.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}