A planning rule based on the Riskiest Assumption Test: list assumptions, rank their risks, and check the most dangerous one before coding. Understanding the request always comes first.
A set of rules for using todotool on tasks that take more than two or three steps. It requires breaking work into small steps, keeping only one step in progress, and updating statuses as work continues.
A recovery rule for when an AI agent stops responding, freezes, or repeats the same actions. It requires finding the likely cause through root-cause analysis before restarting the agent.
A set of rules for checking development work before it is marked complete. It defines success checks such as a clean linter, cross-checking, and challenging the result.
A coding rule that sets maximum recommended and hard line counts for route, service, and component files. When a file exceeds its recommended size, it requires a splitting plan before more code is added.
A set of coding rules for React Hooks, which are reusable functions for managing component behavior. It requires Hooks to run in the same order on every render and before any early exit from the component.
Apply when writing API routes, database queries, form handling, or user input processing. Enforces input validation (Zod), SQL injection prevention, XSS protection. Target: zero security vulnerabilities in new code.
Lets the agent inspect a PostgreSQL database schema and run read-only SQL queries against it. Runs locally from the @modelcontextprotocol/server-postgres npm package. Needs 1 environment variable to run.
AGENTS.md instructions for dmitryprg-ai/cursor-develop-autorules, covering agents.md — crmai project, production services, key files, chat system messages (open lines) and architecture v13.0: dual ide support (cursor + claude code).