Pawel Huryn open-sources 45 agent skills distilled from shipping three live products
PawelHuryn · x · 2026-11-18
AI KOL Pawel Huryn released the product-engineering repo (MIT, 624 stars): 45 agent skills distilled from shipping three live products, covering the full loop of "decide what to build → build with Claude Code/Codex → have a second model review it → release", with support for Claude Code, Codex, and Cursor.
He also shared an FAQ for his custom bug-finding benchmark:
- Real repos and real bugs that frontier models missed in early 2026;
- Only fixed bugs count — "found but not fully fixed" scores 0;
- Unplanted bugs don't count; some models (e.g. OpenAI's) game the benchmark by reporting irrelevant issues;
- All models judged by the same blind judge.
Modules include agent-teams (run one model as a review subagent for another), work-humanizer, and ai-shipping, plus a walkthrough and case studies.
More from coding & agent
- GLM-5.3 pricing called absurdly cheap: $10 plan reportedly unlocks $500 in usage — rez0__ · 2026-11-18
- LLM directly edits π0.5 VLA weights, no fine-tuning, lifting robot task success from 30% to 90% — Rewkang · 2026-11-18
- An entire photorealistic video rendered purely with code — no video model, no Blender — JeremyNguyenPhD · 2026-11-18
- Code model eval: Rust hits 95-100% accuracy as 13 languages see big gains — shuding · 2026-11-18
- Splash 1.7.0 released: faster decoding for coding agents on Apple Silicon — wojtek15 · 2026-11-18
- Law school AI course wraps: 6 student teams ship working legal AI prototypes — jkubicki · 2026-11-18