First open-source project. Learning as I go. That said, I genuinely want collaborators — this started as a personal system and needs other perspectives.
- Port analyze.py to Node.js — The learning loop is ~970 lines of Python. Everything else is Node/shell. One runtime = lower adoption barrier.
- Tests — Zero tests. The promotion logic (auto-writing to
.claude/rules/) especially needs coverage. - Linux/WSL testing — Built on macOS M1, never tested anywhere else.
- Domain packs — Professional expertise + Claude Code experience = installable knowledge. See
domain-packs/example-pack/for the format. - Architecture feedback — Is the config schema sane? Is the hook pipeline right? Would you actually install this? Why not?
git clone https://github.com/somarakis/cortex.git
cd cortex
npm install- Bug reports: Open an issue. Include OS, Node version, Claude Code version.
- Ideas: Open an issue tagged "idea" or "discussion".
- Code: Fork, branch, PR. One thing per PR.
- Domain packs: Copy
domain-packs/example-pack/, fill in your expertise.
Read these in order:
- README.md — What it is and why
- EVOLUTION.md — How it got here (all the wrong turns included)
- docs/hooks/ — Hook lifecycle and authoring guide
- ROADMAP.md — Where it's going
No linter yet. Follow conventions you see in existing code. If something's inconsistent, fix it.
I'm a lawyer who learned to code by building this with Claude Code. My code probably looks like it. Constructive feedback > polite silence. Tell me what doesn't make sense.
By contributing, you agree your work is licensed under MIT.