Install:
npx skills add zning1994/docs-organization
https://t.co/CkbGQvyNfe
Built from lessons reorganizing a 6-agent AI system with 70+ API endpoints and 39 skills.
#ZNing My CLAUDE.md quietly grew to 800 lines.
It loads into every single AI conversation. All of it. Every time.
I got it down to 250 with zero information loss. Here's what I learned:
#AICoding#OpenSource#ClaudeCode
Works across 30+ tools thanks to the Agent Skills open standard:
Claude Code, Codex CLI, Gemini CLI, Cursor, GitHub Copilot, Kimi CLI...
One SKILL.md, runs everywhere.
I read a blog post about Karpathy's autoresearch method. A few hours later, I had an open-source skill for our AI agents on GitHub.
The idea: stop hand-tuning AI prompts by feel. Define what "good" means as binary evals, then let a script do the work — mutate, test, score, keep or discard, repeat.
I built it as a zero-dependency Python script that works with any AI coding agent (OpenClaw, Claude Code, Cursor, Cline). Tested on a real skill: pass rate went from 37.5% to 54.2% in 5 automated experiments. No manual tweaking.
It's open source:
🔗 GitHub: https://t.co/fwNohyNjsB
🔗 ClawHub: https://t.co/EuShL01JZ8
Full write-up in the article below 👇
#PromptEngineering #AI #OpenSource #LLM #Autoresearch #OpenClaw #ClaudeCode #DevTools #AIAgents #ZNing
https://t.co/myJ9Qctmgu
I launched https://t.co/tNYOm7V5wD last night and already 130+ people have signed up including an OF model (lmao) and the CEO of an AI startup.
If your AI agent wants to rent a person to do an IRL task for them its as simple as one MCP call.
Due to laws against data export, Tesla achieved the top results in China despite having no local training data.
Tesla is adding training data from our world simulator and test tracks to achieve 6/6.