/ Каталог / Песочница / agent-rules-skill
● Сообщество netresearch ⚡ Сразу

agent-rules-skill

автор netresearch · netresearch/agent-rules-skill

Generates AGENTS.md files following the agents.md convention — tells any coding agent (Claude Code, Cursor, Codex) how your repo works.

A small, focused Claude Code skill by Netresearch. It inspects a repository and drafts an AGENTS.md file following the agents.md convention — a growing standard for telling AI agents project conventions, build commands, testing approach, and contribution rules in one place.

Зачем использовать

Ключевые функции

Живое демо

Как выглядит на практике

agent-rules-skill.replay ▶ готово
0/0

Установка

Выберите клиент

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "agent-rules-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/netresearch/agent-rules-skill",
        "~/.claude/skills/agent-rules-skill"
      ],
      "_inferred": true
    }
  }
}

Откройте Claude Desktop → Settings → Developer → Edit Config. Перезапустите после сохранения.

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "agent-rules-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/netresearch/agent-rules-skill",
        "~/.claude/skills/agent-rules-skill"
      ],
      "_inferred": true
    }
  }
}

Cursor использует ту же схему mcpServers, что и Claude Desktop. Конфиг проекта приоритетнее глобального.

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "agent-rules-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/netresearch/agent-rules-skill",
        "~/.claude/skills/agent-rules-skill"
      ],
      "_inferred": true
    }
  }
}

Щёлкните значок MCP Servers на боковой панели Cline, затем "Edit Configuration".

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "agent-rules-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/netresearch/agent-rules-skill",
        "~/.claude/skills/agent-rules-skill"
      ],
      "_inferred": true
    }
  }
}

Тот же формат, что и Claude Desktop. Перезапустите Windsurf для применения.

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "agent-rules-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/netresearch/agent-rules-skill",
        "~/.claude/skills/agent-rules-skill"
      ]
    }
  ]
}

Continue использует массив объектов серверов, а не map.

~/.config/zed/settings.json
{
  "context_servers": {
    "agent-rules-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/netresearch/agent-rules-skill",
          "~/.claude/skills/agent-rules-skill"
        ]
      }
    }
  }
}

Добавьте в context_servers. Zed перезагружается автоматически.

claude mcp add agent-rules-skill -- git clone https://github.com/netresearch/agent-rules-skill ~/.claude/skills/agent-rules-skill

Однострочная команда. Проверить: claude mcp list. Удалить: claude mcp remove.

Сценарии использования

Реальные сценарии: agent-rules-skill

Generate an AGENTS.md for a repo that doesn't have one

👤 Teams introducing AI coding agents to an existing codebase ⏱ ~15 min beginner

Когда использовать: Your team just started using Claude Code / Cursor and nobody's written the orientation doc.

Предварительные требования
  • Skill installed — git clone https://github.com/netresearch/agent-rules-skill ~/.claude/skills/agent-rules-skill
Поток
  1. Run it on your repo
    Use agent-rules-skill. Analyze this repo and draft an AGENTS.md following the convention.✓ Скопировано
    → Draft with build/test commands, lint config, style notes, gotchas
  2. Refine
    Add a section on where not to edit (generated files, vendor/).✓ Скопировано
    → Section added with explicit paths
  3. Commit
    Commit as docs/AGENTS.md.✓ Скопировано
    → File committed at repo root or in docs/

Итог: A living AGENTS.md that makes future agent sessions start with context.

Подводные камни
  • AGENTS.md balloons into an everything-doc — Keep it focused on what an agent needs. Link to README for humans.
Сочетать с: filesystem · git

Generate per-package AGENTS.md files in a monorepo

👤 Monorepo maintainers ⏱ ~30 min intermediate

Когда использовать: Different packages have different conventions; a single root AGENTS.md isn't enough.

Поток
  1. Scan packages
    Use agent-rules-skill. For each package in packages/, generate a tailored AGENTS.md capturing its specifics.✓ Скопировано
    → One AGENTS.md per package
  2. Root summary
    Generate a root AGENTS.md that links to per-package files and lists shared rules.✓ Скопировано
    → Root file with pointers

Итог: Agents pick up the right context based on which package they're working in.

Сочетать с: filesystem

Keep AGENTS.md in sync with an evolving repo

👤 Repos where AGENTS.md goes stale fast ⏱ ~10 min beginner

Когда использовать: You added new build commands or changed conventions and AGENTS.md lies about them.

Поток
  1. Diff
    Use agent-rules-skill. Compare current AGENTS.md to what the repo actually has now. What's stale?✓ Скопировано
    → List of drift items
  2. Update
    Update only the stale parts; leave human-written sections alone.✓ Скопировано
    → Minimal diff preserving human commentary

Итог: AGENTS.md stays trustworthy.

Сочетать с: git

Комбинации

Сочетайте с другими MCP — эффект x10

agent-rules-skill + filesystem

Apply generated files directly

Write the draft to AGENTS.md at repo root.✓ Скопировано
agent-rules-skill + git

Commit as a focused change

Commit the new AGENTS.md with a message describing what it covers.✓ Скопировано

Инструменты

Что предоставляет этот MCP

ИнструментВходные данныеКогда вызыватьСтоимость
scan_repo repo path First step on any repo 0
draft_agents_md scan output Drafting new file 0
diff_check existing AGENTS.md + current repo Staleness check 0

Стоимость и лимиты

Во что обходится

Квота API
None
Токенов на вызов
Modest — repo scan + short output
Деньги
Free
Совет
Run on focused directories in huge monorepos; whole-repo scans are token-heavy.

Безопасность

Права, секреты, радиус поражения

Хранение учётных данных: None
Исходящий трафик: Repo contents go through Claude API during scan

Устранение неполадок

Частые ошибки и исправления

Generated commands don't match CI

Verify against actual CI config; skill infers from common files but not always correct

File is too long

Prompt for conciseness; agents.md is meant to be short

Альтернативы

agent-rules-skill в сравнении

АльтернативаКогда использоватьКомпромисс
Manually writing AGENTS.mdSmall repo with clear conventionsWorks but drifts; need discipline to keep fresh
Cursor-specific .cursorrulesYou only use CursorNon-portable; AGENTS.md works across agents

Ещё

Ресурсы

📖 Читать официальный README на GitHub

🐙 Открытые задачи

🔍 Все 400+ MCP-серверов и Skills