interview-battlecard

Interview Battlecard

When to use

Stage 5 of the career-agent workflow, only after the tailored resume and the interview-coach prep doc both exist. The battlecard is a derivative — it does not introduce new content, it compresses what's already in those two files into something the candidate can scan in 60 seconds before walking into a room.

Inputs (read these first)

The main agent passes the resume slug and JD slug as part of the orchestration context. Read all three at limit=1000:

• /tailored_resume/<resume-slug>/<jd-slug>.yaml — for the candidate's strongest claims and headline metrics.
• /interview_coach/<resume-slug>/<jd-slug>.md — for the round taxonomy, STAR stories (use the 60-second versions), questions to ask back, and watch-outs.
• /research/<resume-slug>/<jd-slug>.md — for the 3 punchiest company facts.

Outputs

Two siblings on disk; the JSON is the source of truth (LLM-written, user-editable), the PDF is the day-of artifact the candidate downloads and reviews.

/interview_battlecard/<resume-slug>/<jd-slug>.json
/interview_battlecard/<resume-slug>/<jd-slug>.pdf

Two-step flow, in this exact order:

1. write_file("/interview_battlecard/<resume>/<jd>.json", <json string>) — write the JSON source.
2. render_battlecard_pdf("/interview_battlecard/<resume>/<jd>.json") — render the PDF beside it.

See the "## Updates" section below for follow-up edits (adding a round, swapping a story, re-rendering after a user hand-edit).

JSON shape (strict)

The JSON object must validate against this shape exactly. The rounds array order must match the interview-coach prep doc; do not invent or skip rounds.

{
  "document_title": "Interview Battle Card — <Candidate Name>",
  "rounds": [
    {
      "title": "<Round name> — <format>",
      "subtitle": "<Company> · <Role>",
      "introduction": ["<keyword 1>", "<keyword 2>", "..."],
      "stories_ready": [
        { "title": "<Story title>", "body": "<15-sec one-liner>" }
      ],
      "company_facts": ["<falsifiable fact 1>", "..."],
      "questions": ["<grounded question 1>", "..."],
      "watch_outs": ["<risk + 1-line mitigation 1>", "..."]
    }
  ]
}