interviewstreet/hiring-agent — reverse-engineered prompt

Build me a simple resume scoring tool in Python that I can run from the command line by giving it a PDF path. I want it to read the resume, pull out the main sections like basics, work history, education, skills, projects, and awards, then turn that into clean structured data. If the resume includes a GitHub profile, use that too by looking at the candidate’s profile and repos, picking the most meaningful projects, and using that as extra signal.

The main goal is a fair and explainable evaluation, not just a mystery score. Give category scores for things like open source work, personal projects, production experience, and technical skills, plus any bonus points or deductions, and show the evidence behind the decision in plain English.

Please make it work with either a local model or Gemini, depending on config. I’d also like a development mode that caches intermediate results and appends evaluations to a CSV so I can compare candidates over time. Look up current docs online if you need to.