0x7oid/code.rev — reverse-engineered prompt

Build me a Python command line tool that reviews another Python project in a few passes and writes separate JSON reports into a review_output folder inside the target project.

I want one pass for static code issues, one for security, one for logic and correctness, and one for overall structure and architecture. For the static checks, use normal Python linting and type checking. For security, first run a pattern based scan, then if there are findings, send them to an LLM to judge whether they are actually risky in context. The logic and structure passes should also use an LLM.

The tool should take a path to the project and an optional model name from the command line, keep going even if one phase fails, and use an API_KEY environment variable for the LLM powered steps. If trust_model.md or architecture_intent.md exist in the target project, include them in the review. Look up current docs online if you need to.