- Home
- Skills
- Agents & Orchestration
- CLAUDE.md / AGENTS.md Generator: Onboard Agents to Any Repo
Works with the AI Agents you already use
CLAUDE.md / AGENTS.md Generator: Onboard Agents to Any Repo
- **Verifies every command**: install, dev, build, lint, typecheck, test-all and single-test commands are actually executed.
Free
CLAUDE.md / AGENTS.md Generator: Onboard Agents to Any Repo
Example session with this skill installed
Create a CLAUDE.md for this repo so Claude Code knows how to build, test and follow our conventions. Verify the commands before documenting them.
- Read your context and instructions
- Compiled the claude md agents
Summary
Scanned 214 tracked files, detected a pnpm + Turborepo monorepo (Next.js web app + Fastify API), ran 7 commands and wrote a 92-line CLAUDE.md at the repo root plus a short apps/api/CLAUDE.md.
Command verification
| Command | Result |
|---|---|
pnpm install --frozen-lockfile | ✅ |
pnpm dev | ✅ web :3000, api :4000 |
pnpm build | ✅ |
pnpm lint | ✅ |
pnpm typecheck | ✅ |
pnpm test | ✅ 312 passed |
pnpm --filter api test -- src/orders.test.ts | ✅ |
Generated CLAUDE.md (excerpt)
# Acme Orders — agent guide
Internal order management: Next.js 14 dashboard + Fastify REST API on PostgreSQL (Prisma).
## Commands
- Install: `pnpm install --frozen-lockfile`
- Dev (all): `pnpm dev` — web :3000, api :4000
- Test one file: `pnpm --filter api test -- <path>`
- Before commit: `pnpm lint && pnpm typecheck && pnpm test`
## Architecture
- `apps/web/` — Next.js App Router UI, server actions call the API client
- `apps/api/` — Fastify routes → services → Prisma repositories
- `packages/shared/` — zod schemas shared by web and api (source of truth for types)
## Conventions
- Validate all input with zod schemas from `packages/shared`
- Errors: throw `AppError(code, status)`; never return raw Prisma errors
- Tests colocated as `*.test.ts`, use factories in `test/factories`
- Commits: Conventional Commits (`feat(api): ...`)
## Gotchas
- Run `pnpm db:generate` after editing `schema.prisma`
- `apps/web/src/gen/` is generated — never edit
- Integration tests need Docker (`pnpm db:up`)
Open questions
- Which branch deploys to production (
mainorrelease/*)? Not determinable from CI config.
Next step: review the Conventions section, then commit CLAUDE.md so every agent session starts with it.
Connects securely to your tools. The creator never sees your data.
About this skill
Coding agents are only as good as the context they start with. Without a project guide they guess commands, invent folders, ignore your conventions and burn tokens re-exploring the repo on every session.
This skill turns your repository into a short, verified CLAUDE.md or AGENTS.md that any coding agent reads first, so it works like a teammate who already knows the codebase.
What it does
- Discovers the real stack: languages, frameworks, package manager (from the lockfile), monorepo layout, entry points, env vars, DB and migrations.
- Verifies every command: install, dev, build, lint, typecheck, test-all and single-test commands are actually executed. Only working commands are documented, preferring the exact ones your CI uses.
- Extracts conventions from code, not opinion: naming, folder structure, error handling, import aliases, test style and commit format (from
git log). - Documents gotchas and don'ts: generated files not to edit, required env, slow tests, ports, secrets.
- Updates instead of overwriting: if a CLAUDE.md, AGENTS.md, .cursorrules or copilot-instructions file exists, human-written rules are preserved.
- Monorepo aware: optional short nested guides per package.
Why it is different
- Target under 150 lines: every line earns its tokens.
- Ends with a verification report (✅/❌ per command) and the open questions the code could not answer.
- Works for Claude Code (
CLAUDE.md) and Codex, Cursor, Copilot, Gemini CLI (AGENTS.md). Need both? It writes AGENTS.md and a one-line CLAUDE.md that imports it.
Best for
- Starting Claude Code / Codex on an existing repo
- Onboarding agents to a monorepo
- Refreshing an outdated CLAUDE.md after a migration
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
- Command verification runs your project's commands locally; very slow suites or ones needing unavailable services (Docker, cloud credentials) are reported as unverified instead of executed.
- Cannot document infrastructure, secrets or deployment rules that are not visible in the repo or CI config; these are listed as open questions.
- Conventions are inferred from the current code; inconsistent codebases may need a quick human review of that section.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
This skill has been used 14 times.
Trust & safety
Security scan
Security scan passed 11 days ago
- Free with an Agensi account
- Software
GitProject toolchain
Use your project's own toolchain, such as Node.js with pnpm, Python or Go.
Permissions
Read Files: scans source, configs and CI workflows. Terminal: runs your existing install/build/lint/test commands to verify them. Write Files: creates or updates only CLAUDE.md / AGENTS.md (and nested package guides if requested). No network calls of its own and no external services.