Loading Musebox...
Write a rules file that gives AI coding agents the context they keep missing: stack, commands, structure, conventions and the things they must never do, kept short enough to actually be read.
You're a senior engineer setting up a repository for AI coding agents. Write the project rules file the agent reads at the start of every session, so it stops making the same mistakes. About the project (what it is, who uses it): {{project}} Stack and key dependencies: {{stack}} How to install, run, test, lint and build (the actual commands): {{commands}} Folder structure, and where things belong: {{structure}} Conventions and past mistakes agents keep making: {{conventions_and_mistakes}} File: {{{file: CLAUDE.md, AGENTS.md, Cursor project rules, GitHub Copilot instructions}}} **Deliver the file, in Markdown, with these sections:** **Project:** two or three sentences on what this is. **Commands:** install, dev, test, lint, type check and build, each in a code block. **Structure:** where code, tests, docs and config live, and where new files should go. **Conventions:** naming, patterns to follow, and one existing file to copy from for each common task. **Do not:** a short list of hard rules (for example: never commit secrets, never edit applied migrations, never run destructive commands without asking). **Before finishing a task:** the checks to run. **Known issues:** anything currently broken, so the agent doesn't try to "fix" it unasked. Rules: Under about 150 lines; link to longer docs instead of pasting them in. Every rule should be specific enough to check. Only include commands and facts I gave you; mark gaps as [FILL IN]. Put the rules that prevent the most damage first. Tell me where the file goes for the tool I chose.
