AI Toolbox

CLAUDE.md / AGENTS.md Generator

Draft portable agent instructions, choose import or symlink wiring, and lint the result before copying it into your repo.

100% client-side; project instructions stay in this browser. This generator will separate portable AGENTS.md guidance from Claude-specific CLAUDE.md wiring and lint by section, not by a single health score.

# This project — Agent Instructions

## Working agreement
- Read nearby code and existing conventions before editing.
- Keep changes focused; do not rewrite unrelated files.
- Explain assumptions and flag decisions that need product-owner approval.

## Code quality
- Prefer the project’s existing libraries and patterns over introducing new dependencies.
- Add or update focused tests when behavior changes.
- Do not expose credentials, tokens, or private user data in code, logs, or commits.

## Verification
- Run `npm run lint` after relevant code changes.
- Run `npm test` for changed behavior.
- Run `npm run build` before handing off a release-sensitive change.

## Git hygiene
- Preserve unrelated local changes.
- Use descriptive commits and do not force-push shared branches without approval.

@AGENTS.md

## Claude Code
- Use plan mode before multi-file changes, migrations, dependency updates, or external side effects.
- Treat AGENTS.md as the shared source of truth for repository conventions.
- Keep CLAUDE.md concise; move detailed, on-demand procedures into `.claude/skills/` or docs.