199-biotechnologies/context-priming — reverse-engineered prompt

Build me a Python developer tool called Context Priming. I want it to help AI coding agents start with the right information before they begin working, instead of wasting time reading random files or hitting the context limit later.

The tool should take a user task like “fix the auth bug” and a project folder, then scan useful sources such as code files, memory notes, configs, and git history. It should decide what is actually relevant, understand the bigger goal behind the task, and produce a compact starting context that another coding agent can use right away.

Please include a simple command line interface with commands to gather sources and to create the primed context, with normal text output and JSON output. Also make it usable as a Python library where I can plug in my own LLM call. Add optional support for Claude Code or Claude Agent SDK if that fits cleanly.

Use Python 3.11 or newer, add basic tests, and write clear setup and usage instructions. Look up current docs online if needed.