oritera/Cairn — reverse-engineered prompt

Build me a Python and Docker based tool like Cairn, a general problem solving engine where I can give it a starting point and a goal, then let AI workers search for a path.

I want it to keep a shared board of confirmed facts, possible next intentions, and human hints I can add while it is running. The workers should read the whole board, decide what to try next, run the exploration, and write back what they found. They should not need fixed roles. The system should create tasks at runtime based on the current state.

Please include a server that stores the graph, a dispatcher that schedules work and manages worker containers, and support for common coding agent backends like Claude Code or Codex through configurable API settings. Make it easy to run with Docker Compose, with a config file for keys and endpoints.

Keep the security use case limited to authorized labs, CTFs, and owned targets. Look up current docs online if you need to.