job-hunter

Job Hunter

A careful, honest job-application copilot for Claude. It scans listings (and flags the ghost jobs and scams), scores each one the way employers actually screen, tailors your CV and cover letter to the roles worth your time, has a simulated recruiter read the result cold before you send it, preps you for the interview, and tracks everything - while following two iron rules:

1. It never invents anything about you. Every line it writes comes from your own CV and your own answers.
2. It never applies, messages, or sends anything for you. It prepares; you submit.

You do not need to be a developer. You do not need to know what git is.

It works for whatever stage you're at. A final-year student, a Master's graduate, a PhD heading for industry or academia, and someone with no degree but years of real work each need a different CV built from the same honest facts - a one-pager led by education and projects, a two-pager led by experience, or a full academic CV with the complete publication list. /setup works out your stage from your own CV, confirms it with you, and every later step follows the right conventions (UK format) automatically.

Install (Claude Desktop - Cowork)

1. Install Claude Desktop and open the Cowork tab.
2. Pick (or create) an empty folder for your job hunt - e.g. My Job Hunt.
3. Go to Customize -> Plugins, add this marketplace, and install job-hunter: zeinali72/job-hunter-plugin
4. In the chat, type /setup and follow along. Have your current CV file ready.

That's it. Your CV, your job list, and everything the plugin writes stay in your folder on your computer.

Install (Claude Code)

/plugin marketplace add zeinali72/job-hunter-plugin
/plugin install job-hunter@job-hunter-marketplace

Then run /setup in your job-hunt folder.

How the pieces fit together

/setup                        once - builds profile/ from your CV and your answers
   |
   v
/connect-linkedin              optional, anytime - lets /scan include your own LinkedIn
   |
   v
/scan <query>              --> jobs/<company>-<role>/job.md   (one per new listing,
   |                           with freshness + ghost-job + scam flags)
   v
/score <listing or "all">  --> jobs/<listing>/score.md   (knockout check + ATS score
   |                           + 10-second skim + A-F fit grade)
   +-- knockout BLOCKED ------------------> skip (no CV fixes a visa/location knockout)
   +-- below threshold, weak fit --------> skip (reported, nothing further)
   +-- below threshold, strong fit (B+) -> "manual-route candidate" (warm intro beats the portal)
   |
   v  ATS >= ats_threshold (default 75, in profile/profile.yml)
/tailor <listing>  +  /cover <listing>  --> cv.md, cover.md   (independent of each other)
   |
   v
/review <listing>          --> jobs/<listing>/review.md   (a recruiter who did NOT write
   |                           it reads it cold + fact-checks every claim - fix, recheck)
   v
/applypack <listing>       --> output/<company>-<role>/   (files + checklist - YOU submit)
   |
   v
/track <listing> <status>  --> jobs/tracker.csv    (log "applied" once you actually submit)
   |
   +--> /outreach <listing>    finds the warm path to the team and drafts the notes
   +--> /interview <listing>   prep pack + live mock when the invite lands
   |
   v
/followup                  --> chases anything "applied" 7+ days ago with no reply

Run each command yourself for full control - or run /pipeline <query> to chain scan -> score -> tailor -> cover -> review -> applypack -> track in one go, stopping automatically at the score gate and the review gate for anything not worth sending. /pipeline is a shortcut, not a replacement: every command below still works standalone, any time, for rerunning a stage or overriding a gate by hand.

The commands