llmd

Today I’m shipping llmd — a CLI that lets you speak human and run shell. Describe what you want, get a verified command, safety warnings for destructive ops, and run with one keystroke. No more googling flags at 2 AM. What makes it different: - Multi‑provider: OpenAI, Claude, Groq, Gemini, OpenRouter. - Self‑verification + confidence scores. - Dangerous command detection with explicit confirmations. - Cross‑platform and local config. Quick tastes: - “create a new branch called feature-auth” → git checkout -b feature-auth - “find all .log files recursively” → find . -type f -name “*.log” - “show processes using >100MB RAM” → ps aux | awk ‘NR==1 || $6>102400’ Getting started: npm install -g llmd-cli llmd setup llmd “list files sorted by size” I’d love feedback on two things: 1) where safety checks feel too strict or too lax, 2) provider/model choices that work best for you. If you try it, tell me what broke so I can fix it before it breaks you.