Skip to content
Loop Engineering
Loop Engineering Tools

AGENTS.md Generator for AI Coding Agents

Create durable AGENTS.md instructions with project rules, exact validation commands, forbidden actions, and a clear definition of done for repeated coding-agent work.

Create Durable Loop Engineering Instructions

Fill in your commands, structure, rules, and done conditions. The AGENTS.md is generated in your browser — copy it into your repository root so every agent run starts from the same stable instructions. This format is summarized from official provider guidance and kept vendor-neutral.

Compare durable instructions in the AGENTS.md, SKILL.md, and RALPH.md guide and verify the finished file with the Loop Engineering checklist.

Copy Your AGENTS.md

AGENTS.md
# AGENTS.md
 
## Project Overview
Loop Engineering is a Next.js + TypeScript + Tailwind CSS project.
 
## Setup
- Install dependencies: pnpm install
- Start development server: pnpm dev
 
## Validation Commands
Run these before considering work complete:
 
- Build: pnpm build
- Test: pnpm test
- Lint: pnpm lint
- Typecheck: pnpm tsc --noEmit
 
## Project Structure
app/ — App Router pages and routes
components/ — reusable UI components
data/ — local TypeScript data files
lib/ — generators, scoring, and helpers
 
## Coding Rules
Prefer small, focused changes.
Reuse existing patterns and components.
Keep the architecture simple.
 
## Forbidden Actions
Do not add a database, authentication, or backend API.
Do not execute user code.
Do not delete or weaken tests to make checks pass.
Do not modify unrelated files.
 
## Review Rules
Review for correctness, safety, and accessibility.
Every blocking issue cites a file and line.
 
## Done When
pnpm lint passes.
pnpm build passes.
All required pages render.
 
## Agent Working Agreement
- Make small, focused changes.
- Prefer existing patterns over new abstractions.
- Do not introduce new dependencies without a clear reason.
- Do not modify unrelated files.
- Do not delete or weaken tests to make checks pass.
- If blocked, stop and summarize the blocker, attempted fixes, and recommended next action.

Good AGENTS.md Checklist

  • Short enough for agents to read
  • Contains exact validation commands
  • Defines forbidden actions
  • Explains project structure
  • Defines what “done” means
  • Avoids long historical notes
  • Avoids outdated project history