Job Hunt Automation Bundle

Job Hunt Automation Bundle is five connected Claude Code skills that run a full-time job search end to end: job-search sources ~20 matching roles for today's target region, scores and filters them, and hands the best 10 off for material prep. job-application-preparation screens each JD the way a hiring manager would — ranks and categorizes the 12-20 most critical keywords, explains what each category signals and why its absence is a red flag, and reports the gap between the JD's must-haves and the candidate's real background. It writes a tailored cover letter and a fully tailored .docx resume, then runs a hybrid ATS score on the draft — a deterministic keyword-matching script plus a dedicated ats-resume-scorer subagent, cross-checked against each other — patches any literal-wording gaps automatically, and logs the result to an applications tracker. job-apply submits every drafted application across any ATS (LinkedIn, Workday, Greenhouse, Lever, Taleo, SuccessFactors, and more), answering standard fields automatically and prompting you only for what it can't answer on its own. job-hunt-setup and job-hunt-update handle onboarding and ongoing changes entirely by chat — no file editing, no terminal, for non-technical users.

Problem it solves: A real job search means re-reading every JD, hand-tailoring a resume and cover letter to match its language, filling out the same 40 form fields on a dozen different ATS platforms, and keeping track of what you sent where and what happened next. Done properly, that's hours per application. This bundle keeps a single master resume and a single set of standing answers as the source of truth, then automates the repetitive parts — sourcing, tailoring, submitting, and tracking — while still stopping to ask you before anything genuinely needs a human judgment call.

Why use this skill: Unlike a single "write me a cover letter" prompt, this is a coordinated pipeline: every skill reads the same master resume and standing-answers file, so your target roles, tone, and disclosed facts stay consistent across every application without re-entering anything. The ATS check is a hybrid, not a single pass — a fast deterministic keyword script catches literal ATS-parser behavior with zero drift between runs, and a separate LLM subagent catches semantic near-misses and wording gaps the keyword layer structurally can't see. The two are reconciled and any real wording gap on genuine experience gets patched into the resume automatically, before it's logged — and any newly surfaced genuine skill gets folded back into your master resume so future tailoring passes never miss it again. It ships as a template with zero personal data baked in — a chat-driven setup skill collects your details conversationally and scaffolds everything for you, including reading an existing resume file to draft the master resume automatically.

Known limitations: the core pipeline needs a browser-automation MCP installed (Claude Code Chrome extension or Playwright MCP) — it does not work with zero MCP servers configured. Setup scripts (setup.sh / setup.ps1) are macOS/Linux/Windows shell scripts — on another agent or a sandboxed environment without shell access, install the skills/*/SKILL.md files and agents/ats-resume-scorer.md by hand. Layer 2 of the hybrid ATS score (ats-resume-scorer) is a Claude Code custom subagent — on another agent, either port it to that agent's subagent format or run with Layer 1 (the keyword script) alone. Job-board coverage and ATS portal support are tuned for the regions and boards named in references/job-boards.md; buyers should review and edit that file for their own target market.