A safe, starter AGENTS.md for your project. It bootstraps a deterministic workflow
that you and your agents can easily extend.
AGENTS.md is a convention file placed at the root of a project that defines rules,
conventions, and session learnings for AI agents working on the codebase. Think of it as
the agent equivalent of CONTRIBUTING.md — it tells agents how to behave, what patterns
are forbidden, and how the project is structured.
This starter file provides:
- Security guards — forbidden patterns (no
eval, no hardcoded secrets, OWASP Top 10 awareness). - Coding standards — an example of Python style, FastAPI conventions, LangGraph patterns. Replace with your language/toolchain.
- Git workflow — conventional commits, no force-push, no rebase without agreement.
- Operational rules — 100% coverage, pre-commit hooks, mocking guidelines.
Edit AGENTS.md directly. Add sections, tighten rules, or bake in project-specific
conventions. Agents reading the file will follow whatever is written there.
Replace the language-specific examples (sections 1, 3–4, 6, 8) with the conventions for your project's stack. Keep the security rules, git conventions, and operational rules as a solid default — they apply regardless of language.
| Section | Purpose |
|---|---|
1. Core Rules |
Forbidden patterns, security mandates, core principles (DRY, KISS, YAGNI) |
2. Project Context |
Description, layout, quick commands |
3. Python Conventions |
Example — tooling, style, error handling, async, testing (adapt for your language) |
4. Framework Conventions |
Example — FastAPI, LangGraph, auth modes (adapt for your frameworks) |
5. Git Conventions |
Commits, branching, review |
6. Operational Rules |
Coverage, pre-commit, mocking gotchas |
7. Session Learnings |
Codebase-specific discoveries |
8. Checklist |
Example — Python-specific type annotations, pylint, coverage, mypy markers (adapt for your stack) |
