← ClaudeAtlas

jd-keyword-minerlisted

Mines keywords, skills, and exact ATS phrases from one or more LinkedIn job posts (or pasted job descriptions) and writes a keyword report. Use when the user says "mine this job post", "keywords from this JD", "what does this role want", "analyze these job postings", or runs /jd-keyword-miner. To then rewrite a profile or resume with these keywords, use the rewrite-linkedin-profile or rewrite-resume skills.
stalinzbb/job-search-toolkit · ★ 0 · Data & Documents · score 58
Install: claude install-skill stalinzbb/job-search-toolkit
# JD Keyword Miner Turn job posts into a keyword report the user can write against. Multiple posts for the same kind of role is the point — repetition across posts is what separates a real requirement from one company's wishlist. ## Output location Everything for one target role lives in one folder: ``` $JOB_SEARCH_DIR/<company>-<role-slug>/ ├── inputs/ # raw job posts, the resume you provided ├── keyword-report-2026-08-24.md # every output is dated; newest wins ├── linkedin-rewrite-2026-08-24.md └── resume-rewrite-2026-08-24.md ``` `JOB_SEARCH_DIR` is the environment variable if the user has set one, otherwise `~/job-search`. Resolve it once with Bash — `echo "${JOB_SEARCH_DIR:-$HOME/job-search}"` — and never hardcode a path. Create the folder on first run and tell the user where it went. The newest report is whichever `keyword-report-*.md` sorts last: `ls "$DIR"/keyword-report-*.md | tail -1`. Reports are dated by run (`keyword-report-<YYYY-MM-DD>.md`; a second run the same day gets a `-2` suffix). Never overwrite an existing report — every run adds one, so the folder becomes a history you can diff against. Raw job text goes to `inputs/jd-<n>.md`, numbered continuing across runs (if `jd-4.md` exists, the next post is `jd-5.md`). Pick the slug from the dominant company+role (e.g. `stripe-senior-product-designer`). For a multi-company cluster with no single employer, use `<role-slug>` alone (e.g. `senior-product-designer`) and say so. If a fol