(did this a lot before it was native in product) to gather relevant context and align on the problem/solution with model before executing
2. prompt for self-verifiable solutions the agent can run. if agent can see and reason about what it does, it can detect divergence from plan or what you prompted for in the first place. browser use is the clearest example of this. however, it can also create tests/scripts to verify runtime in all sorts of systems
3. use github cli a lot. with this, agent can see old PR implementation, check action status and logs, get PR comments to address etc. just to verify what's been done previously, but also automating some of the things i would've done by going to github website
the best rules in @cursor_ai aren’t the ones you plan. it's the ones you stumble into doing something manually, realizing it’s repetitive, and capturing it as a rule
here’s a overview of the 3 main types of rules i've seen (and use myself) ↓
1. workflows
step-by-step instructions for how cursor should behave in specific situations. written in natural language
get a lot of questions on how to use @cursor_ai in large codebases and monorepos. while it depends™, here are some tips and tricks
enable project structure to make the model more aware of how the project is structured
make sure to @ mention specific files or folders that you know include relevant patterns and examples to the task you want to complete. this grounds the model with your codebase
i've built 10+ internal tools at @cursor_ai using cursor. when you have a platform in place, it makes it so easy to spin up new tools
here's how ↓
the first piece is to have the primitives in place like database, servers, networking, workers etc
i've setup this on aws with terraform
second step is to make sure agent is aware of all primitives. here's a cursor rules i've defined to make it aware of what it can use. this outlines building blocks, references examples and points to MCP tools for introspection
you can now generate rules from conversation using the command /Generate Cursor Rules
great way to capture decisions from back and forth chats you might want to reuse later
we've also made two other important changes to rules:
- agent can now reliably edit them by prompting it
- rules with `Always` attach are no longer forgotten in long chats