claude-code-ops: three AI agents that run your project's ops loop. One sets strategy, one executes, one tracks data. You steer, they work.
I used it to promote another open source project. My total contribution: 30 minutes setting direction, picking 3 drafts. The rest -- research, writing, scoring, prioritizing -- was the system.
How it works:
- Conversation Director: chat with it to set strategy. It works through tasks, turns successful actions into reusable skills.
- Auto Director: runs unattended. Reviews results, assigns tasks, extracts learnings, writes daily reports.
- Specialist: executes per skill. Creates content, collects data, scans intel.
All state lives in files, not in conversations. Close the terminal, open it tomorrow, pick up from where you left off. One command to start: bash run.sh.
Ships with zero domain skills. Use it once, it learns once. You figure out what works, it becomes a reusable skill.
Honest bits:
- Currently used by one person (me).
- Requires basic command line and Claude Code knowledge.
- Not plug-and-play -- you spend 30 minutes setting strategy first, then it works.
- Publishing has a safety lock. Nothing goes out without your say-so.
Open source, built on Claude Code:
Top comments (0)