- 🔒 AI-agent security — SkillJail, a kernel-enforced least-privilege sandbox for AI-agent skills, and skill-scanner, a read-only, offline scanner that tells you whether a skill is safe to install before it touches your machine.
- 🤖 Automation & ML — pipelines that do real work end-to-end, from OSS contribution to applied ML.
- 🌍 Open source — I contribute fixes upstream across the ecosystem (see the auto-updating list below).
Repositories where I've had pull requests merged. This list rebuilds itself via GitHub Actions as new PRs land.
- puppeteer/puppeteer · 2×
- borgbackup/borg
- datachain-ai/datachain
- duplicati/duplicati
- github/advisory-database
- gyroflow/gyroflow
- jaegertracing/jaeger
- kornia/kornia
- nextlevelbuilder/ui-ux-pro-max-skill
- noctalia-dev/noctalia
- numpy/numpy
- openai/codex-security
- sgl-project/sglang
🤖 This README is partly self-updating — a scheduled GitHub Action refreshes the stats and the contributions list as my work gets merged.