job-hunter

Execute `.claude/skills/caveman/SKILL.md` inline before processing any command. Execute `.claude/skills/job-hunter/_rules.md` inline before processing any command. # Job Hunter Command Center One entry point for all job search work. Keep output compact — no raw candidate snapshots, full logs, PDFs, or Draft/raw story-bank sections. Arguments: `$ARGUMENTS` ## Key Paths | What | Path | |---|---| | Story bank | `profile/story_bank.md` | | Candidate queue | `outputs/state/agent_candidate_queue.json` | | Job outputs | `outputs/jobs/<slug>/` | | Job DB | `outputs/state/jobs.db` | | Config | `config/job_hunter.yml` | ## Safety Rules - Never fabricate resume facts, employers, dates, metrics, skills, or job history. - Never submit applications, send messages, connect, follow, like, comment, or post automatically. - Never overwrite profile/config inputs unless that mode requires it and the user requested it. - Leave generated changes uncommitted unless the user runs `job-hunter finalize`. ## Initialization Before any mode that reads profile files (batch, one, tailor, score, screen, interview, outreach), run: ```bash job-hunter internal compile-profile ``` This compiles profile files into minified versions for the session. A missing config or profile file is a silent no-op (fine for a fresh workspace). Any other failure logs one compact warning line to `job_hunt.log` and falls back to the raw profile files — it never raises or blocks the session. ## Routing Normalize the fi...