{{PROJECT_NAME}}
Instructions for any AI coding agent working on this project.
This is the primary instructions file.
AGENTS.mdis the emerging universal standard - supported natively by Codex, Cursor, Windsurf, Aider, and read by Claude Code as a fallback whenCLAUDE.mddoesn't exist. For GitHub Copilot, copy the reference table below into.github/copilot-instructions.md. For ChatGPT (no native support), paste the relevantdocs/ai/file into your conversation.
Tech Stack
{{TECH_STACK_DESCRIPTION}}
Hard Rules
Writing
- (Add project writing conventions here, e.g. "Never use em dashes.")
Code
- (Add universal code rules here, e.g. "Comment the 'why', not the 'what'.")
- (e.g. "SEARCH before you write - grep the codebase for existing implementations before adding new helpers.")
- (e.g. "One source, one truth - extract shared logic, never copy-paste patterns.")
Commands
# Add the most important commands here, e.g.
# npm run dev
# npm test
# npm run build
Reference Table
When you're about to work on something, read the matching docs/ai/ file BEFORE diving in:
| Working on... | Read |
|---|---|
| Editing these instructions or skills | docs/ai/writing-docs.md |
| {{REFERENCE_TABLE_ROWS}} |
Multiple rows often apply. When in doubt, read more rather than less.
Context Loading
When starting work on a task, state which docs/ai/ files you're reading before diving in, in this format:
📖 Loading context: docs/ai/<file>.md, docs/ai/<other>.md
This lets the user verify the right knowledge loaded. If no docs/ai/ files are relevant, say so.
Architecture
This project uses the three-layer adaptive docs system - see docs/ai/writing-docs.md for the full explanation.
- Layer 1 (this file) - universal rules + reference table. Always loaded.
- Layer 2 - skills in
.agents/skills/(auto-activated by description matching) and nestedAGENTS.mdfiles in source directories (auto-loaded when editing those directories). - Layer 3 - reference docs in
docs/ai/(loaded on-demand via the table above).