
Every AI coding tool starts from zero context.
You open Claude Code or Cursor, and the AI has no idea what your project does, which patterns you follow, or what "good" looks like in your domain.
So you explain it. Every. Single. Session.
There's a better way.
Both Claude Code and Cursor support project-level config files:
.claude/rules/ — Claude Code reads every .md file here at session start
.cursor/rules/ — Cursor reads every file here automatically
These are native features. No extensions. No API keys. Most developers don't know they exist.
Rules — always-on instructions that shape every response:
"Always use type hints in Python"
"Follow the Airbnb style guide for JavaScript"
"Think about security implications before suggesting code"
Skills — domain knowledge docs that load when relevant:
Your API schema
Your deployment process
Your testing conventions
Instead of writing these configs from scratch, I built a library of 38 pre-configured personas for different dev roles:
PersonaWhat it doessenior-python-developerType hints, FastAPI patterns, pytestsecurity-engineerThreat modeling, OWASP, blast radiusdevops-cloud-engineerTerraform, CI/CD, observabilitysenior-frontend-developerReact, TypeScript, a11ysystem-architectService boundaries, tradeoffs...33 more roles
Clone the repo: git clone https://github.com/ratnesh-maurya/cursor-claude-personas
Copy the persona folder into your project: cp -R senior-python-developer/.claude/ your-project/
Reload your editor.
That's it. The AI reads the configs and behaves like a domain specialist.
Fork the repo, pick the closest persona, and edit the rules to match your team's conventions. The structure is simple:
.claude/ rules/ your-rule.md # Always-on behavior skills/ your-skill/ SKILL.md # Domain knowledge
MIT licensed: https://github.com/ratnesh-maurya/cursor-claude-personas
Star it if it's useful. PRs welcome.
0
1
0