29k-Star RAG Maintainer Open-Sources Kit That Makes Agents Prove Docs Match Code
Nir777 · reddit · 2026-09-15
Nir Diamant, maintainer of the 29k-star RAG Techniques repo, open-sourced (Apache 2.0) the kit he uses to verify agent memory and docs. Asking Claude five questions before reading code revealed it marked wrong answers SURE, mistaking planned tools in a spec for the actual stack. His kit generates a docs layer and CLAUDE.md from real code, flags anything unverifiable as unconfirmed, and caught five README/spec claims the code disproves. Pasting one line clones the repo and runs RUN.md; it deletes nothing. He also benchmarked Claude Code's /init, which caught a migration-ordering landmine his kit missed. Limits: nothing enforces the files, and the final check is one model grading another—worse in Codex/Cursor without sub-agents.
More from coding & agent
- Voice agent Fo ships major voice stack update: call it, and it calls for you — Scobleizer · 2026-09-15
- 25-year veteran runs Opus+Codex Astra adversarial coding loop at $400/month — Due_Warthog749 · 2026-09-15
- Liquid AI open-sources 'antidoom' FTPO training to fix small-model doom loops — helloiamleonie · 2026-09-15
- AI coding tools keep leaking conversation history into commit messages and comments — recro69 · 2026-09-15
- Sandbox Map for Claude Code and Codex Agents: From Docker SBX to E2B — minchoi · 2026-09-15
- Software costs have collapsed: we are all Product Engineers now — rseroter · 2026-09-15