langchain-ai/openwiki — reverse-engineered prompt
Reverse engineered prompt
Build me a command line tool called OpenWiki that can scan a code repository and write useful documentation for coding agents, then keep that documentation fresh over time. I want an easy setup flow where running an init command walks me through choosing an AI provider, entering an API key, picking a model, and optionally adding LangSmith tracing, then saves that local config for later.
The tool should create an openwiki folder when no docs exist yet, and if that folder is already there it should update the docs based on recent repo changes. I also want an interactive mode where I can chat with it, plus a simple one shot mode where I pass a prompt and it prints the result and exits. It should also be able to add or update AGENTS.md and CLAUDE.md so other coding agents know to use the generated docs for context.
If it makes sense, include an example GitHub workflow that can open a daily PR with doc updates. Look up current docs online if you need to.
Want more depth? Deep Reverse