agents-md-generator
Scannednpx machina-cli add skill uts58/my-random-skills/agents-md-generator --openclawAGENTS.md Generator
Generate hierarchical AGENTS.md structures optimized for AI coding agents with minimal token usage.
Core Principles
- Root AGENTS.md is LIGHTWEIGHT - Only universal guidance, links to sub-files (~100-200 lines max)
- Nearest-wins hierarchy - Agents read closest AGENTS.md to file being edited
- JIT indexing - Provide paths/globs/commands, NOT full content
- Token efficiency - Small, actionable guidance over encyclopedic docs
- Sub-folder files have MORE detail - Specific patterns, examples, commands
Workflow
Phase 1: Repository Analysis
Analyze and report:
- Repository type: Monorepo, multi-package, or simple?
- Tech stack: Languages, frameworks, key tools
- Major directories needing own AGENTS.md:
- Apps (
apps/web,apps/api,apps/mobile) - Services (
services/auth,services/transcribe) - Packages (
packages/ui,packages/shared) - Workers (
workers/queue,workers/cron)
- Apps (
- Build system: pnpm/npm/yarn workspaces? Turborepo? Lerna?
- Testing setup: Jest, Vitest, Playwright, pytest?
- Key patterns: Organization, conventions, examples, anti-patterns
Present as structured map before generating files.
Phase 2: Root AGENTS.md
Create lightweight root (~100-200 lines):
# Project Name
## Project Snapshot
[3-5 lines: repo type, tech stack, note about sub-AGENTS.md files]
## Root Setup Commands
[5-10 lines: install, build all, typecheck all, test all]
## Universal Conventions
[5-10 lines: code style, commit format, branch strategy, PR requirements]
## Security & Secrets
[3-5 lines: never commit tokens, .env patterns, PII handling]
## JIT Index
### Package Structure
- Web UI: `apps/web/` -> [see apps/web/AGENTS.md](apps/web/AGENTS.md)
- API: `apps/api/` -> [see apps/api/AGENTS.md](apps/api/AGENTS.md)
### Quick Find Commands
- Search function: `rg -n "functionName" apps/** packages/**`
- Find component: `rg -n "export.*ComponentName" apps/web/src`
- Find API routes: `rg -n "export const (GET|POST)" apps/api`
## Definition of Done
[3-5 lines: what must pass before PR]
Phase 3: Sub-Folder AGENTS.md
For each major package, create detailed AGENTS.md:
# Package Name
## Package Identity
[2-3 lines: what it does, primary tech]
## Setup & Run
[5-10 lines: install, dev, build, test, lint commands]
## Patterns & Conventions
[10-20 lines - MOST IMPORTANT SECTION]
- File organization rules
- Naming conventions
- Examples with actual file paths:
- DO: Use pattern from `src/components/Button.tsx`
- DON'T: Class components like `src/legacy/OldButton.tsx`
- Forms: Copy `src/components/forms/ContactForm.tsx`
- API calls: See `src/hooks/useUser.ts`
## Key Files
[5-10 lines: important files to understand package]
- Auth: `src/auth/provider.tsx`
- API client: `src/lib/api.ts`
- Types: `src/types/index.ts`
## JIT Index Hints
[5-10 lines: search commands for this package]
- Find component: `rg -n "export function .*" src/components`
- Find hook: `rg -n "export const use" src/hooks`
- Find tests: `find . -name "*.test.ts"`
## Common Gotchas
[3-5 lines if applicable]
- "Auth requires NEXT_PUBLIC_ prefix for client-side"
- "Always use @/ imports for absolute paths"
## Pre-PR Checks
[2-3 lines: copy-paste command]
pnpm --filter @repo/web typecheck && pnpm --filter @repo/web test
Phase 4: Special Templates
Design System / UI Package
## Design System
- Components: `packages/ui/src/components/**`
- Use design tokens from `packages/ui/src/tokens.ts`
- Component gallery: `pnpm --filter @repo/ui storybook`
- Examples:
- Buttons: `packages/ui/src/components/Button/Button.tsx`
- Forms: `packages/ui/src/components/Input/Input.tsx`
Database / Data Layer
## Database
- ORM: Prisma / Drizzle / TypeORM
- Schema: `prisma/schema.prisma`
- Migrations: `pnpm db:migrate`
- Connection: via `src/lib/db.ts` singleton
- NEVER run migrations in tests
API / Backend Service
## API Patterns
- REST routes: `src/routes/**/*.ts`
- Auth middleware: `src/middleware/auth.ts`
- Validation: Zod schemas in `src/schemas/**`
- Errors: `ApiError` from `src/lib/errors.ts`
- Example: `src/routes/users/get.ts`
Testing
## Testing
- Unit: `*.test.ts` colocated
- Integration: `tests/integration/**`
- E2E: `tests/e2e/**` (Playwright)
- Single test: `pnpm test -- path/to/file.test.ts`
- Mocks: `src/test/mocks/**`
Output Format
Provide files in order:
- Analysis Summary
- Root AGENTS.md (complete)
- Each Sub-Folder AGENTS.md (with file path)
Format:
---
File: `AGENTS.md` (root)
---
[content]
---
File: `apps/web/AGENTS.md`
---
[content]
Quality Checklist
Before generating, verify:
- Root AGENTS.md under 200 lines
- Root links to all sub-AGENTS.md files
- Each sub-file has concrete examples (actual paths)
- Commands are copy-paste ready
- No duplication between root and sub-files
- JIT hints use actual patterns (ripgrep, find, glob)
- Every "DO" has real file example
- Every "DON'T" references real anti-pattern
- Pre-PR checks are single commands
Source
git clone https://github.com/uts58/my-random-skills/blob/main/agents-md-generator/SKILL.mdView on GitHub Overview
Generates lightweight AGENTS.md structures optimized for AI coding agents. It analyzes the repository to propose a Root AGENTS.md and per-folder sub-files, using a nearest-wins hierarchy and JIT indexing, keeping documentation tiny yet actionable.
How This Skill Works
Phase 1 analyzes repository type, tech stack, and major directories to map where AGENTS.md is needed. Phase 2 generates a lightweight Root AGENTS.md of about 100-200 lines with setup commands, universal conventions, and security notes. Phase 3 creates detailed subfolder AGENTS.md for each major package with file patterns, examples, and JIT search hints. The result is token-efficient guidance that directs AI agents without embedding full code.
When to Use It
- Create AGENTS.md files across a codebase to standardize AI-friendly docs
- Analyze a repository to determine which directories need AGENTS.md
- Set up AI-friendly project documentation with concise guidance
- Generate context files for AI coding assistants to read during edits
- Respond to triggers like create AGENTS.md, generate agents, or hierarchical agents
Quick Start
- Step 1: Run the AGENTS.md generator on your repository
- Step 2: Review Phase 1 output and tailor sections for your repo
- Step 3: Commit and link root and subfolder AGENTS.md files
Best Practices
- Begin with a lightweight root AGENTS.md (~100-200 lines) linking to sub-files
- Keep sub-folder AGENTS.md detailed where it matters most
- Provide concrete file paths, globs, and commands instead of full content
- Use a nearest-wins approach to minimize documentation drift
- Review and update AGENTS.md as the codebase evolves and add new dirs
Example Use Cases
- Onboarding a new repo by generating root and per-package AGENTS.md
- Monorepo with apps, services, and packages gets hierarchical docs
- AI-assisted documentation setup for legacy projects
- Project-wide context files generated for AI coding assistants
- AGENTS.md scaffolding integrated into PR checks