# Jumping Jobs

> A job-search and resume toolkit that runs inside AI coding agents. Install it once and
> your agent gains seven commands for finding roles, scoring fit, tailoring your resume
> and preparing for interviews.

Jumping Jobs is a skills collection, not a hosted service. It installs into the agent you
already use and keeps everything it produces in your own repository as plain Markdown and
HTML. Free and open source under the MIT licence. Requires Node 18 or newer.

## Install

- Install · auto-detects your harness: `npx jumpingjobs install`
- 1 · Add the marketplace: `/plugin marketplace add jumpingjobs/jumpingjobs`
- 2 · Install the plugin: `/plugin install jumpingjobs@jumpingjobs`
- Install · one harness: `npx jumpingjobs install --harness cursor`

Then run `/init-resume` in your agent chat to build your profile. Run
`npx jumpingjobs list` to see every supported harness, and `npx jumpingjobs update` to
refresh the skills later.

## Supported harnesses

- Claude Code
- Cursor
- GitHub Copilot
- Gemini CLI
- Codex CLI
- OpenCode
- Kiro

## Commands

- `/init-resume` — **Start here.** Scaffolds your workspace, adds the resume template and builds profile.md by reading your existing resume and interviewing you.
- `/find-jobs` — **Sweep your boards.** Searches your configured job boards for new, on-profile roles, deduped against everything you have saved. Flags roles that are a step up from your current one, and repairs a board's broken extractor instead of reporting it as empty.
- `/scrape-job` — **Save a posting.** Turns a posting URL into structured Markdown, checking the whole tree first so nothing is scraped twice.
- `/assess-job` — **Score the fit.** Rates a posting against your profile and hard constraints. Pros, cons, dealbreaker check, trajectory — step up, lateral or step down — and an honest verdict.
- `/tune-resume` — **Tailor the resume.** Adjusts your resume to the posting. Two pages, honest, verified by page count before it reports done.
- `/interview-prep` — **Research the company.** Deep dossier on company, financials, leadership and competitors, plus a condensed cheat sheet.
- `/interview-debrief` — **Capture the round.** Run right after each interview: verbatim questions, what landed or stumbled, signals, the promised next step. Stumbles become next-round prep.
- `/move-job` — **Track the pipeline.** Moves a posting through the numbered stages — scraped, applied, interview, lost, archived, won — taking its whole file family (resume, cheat sheet, cover letter) along.
- `/assess-offer` — **Evaluate the offer.** Extracts every term from the letter, flags what's missing or ambiguous, checks the paper against what was promised verbally, and drafts your negotiation or response.

Two supporting skills run underneath: `applicant-profile` keeps `profile.md` current,
and `interview-cheatsheet` writes the in-the-room sheet.

## Typical pipeline

1. `/init-resume` (Onboard) — Hand it your resume or LinkedIn; it interviews you for the rest.
2. `/find-jobs` (Sweep) — New roles from your boards, hard constraints flagged not hidden.
3. `/assess-job` (Score) — Fit out of 10 with an honest verdict.
4. `/tune-resume` (Tailor) — A two-page resume saved next to the posting.
5. `/interview-prep` (Prepare) — Dossier plus cheat sheet the night before.
6. `/interview-debrief` (Debrief) — Drain the round into a file before it blurs; stumbles become fixes.
7. `/assess-offer` (Decide) — Every term extracted and weighed; the response drafted, never sent.
8. `/move-job` (Track) — Every stage move takes the posting's whole file family with it.

## How it works

Everything reads one file, `profile.md`, which holds your real metrics, stories and
constraints. Each tailored resume is a two-page projection of that profile, saved next to
the posting it was written for. Anything new you tell the agent is folded back into the
profile so it does not go stale.

Your workspace looks like this:

```
resume/<your-slug>/
  profile.md               # source of truth
  <your-slug>-resume.html  # layout skeleton
  job-boards.md            # boards + search config
  job-postings/            # numbered = pipeline order
    1-scraped/    2-applied/    3-interview/
    4-lost/       5-archived/   6-won/
```

A workspace holds as many applicants as you like — each gets their own folder, profile and
pipeline, so one repository can serve a whole household.

## Principles

- **Approval before it writes.** Resume changes are proposed as an alignment report first. Nothing is written until you say so.
- **Your data stays in your repo.** profile.md, postings and resumes live in your workspace as plain Markdown and HTML. Nothing is uploaded to us.
- **One source, every harness.** Skills are written once and built for each tool. CI fails if a harness tree drifts from source.

## Links

- [Website](https://jumpingjobs.com/)
- [Source and readme](https://github.com/jumpingjobs/jumpingjobs)
- [Quick start](https://github.com/jumpingjobs/jumpingjobs#quick-start)
- [Adding a harness](https://github.com/jumpingjobs/jumpingjobs/blob/main/docs/DEVELOP.md)
- [Resume templates](https://github.com/jumpingjobs/jumpingjobs/tree/main/templates)
- [npm package](https://www.npmjs.com/package/jumpingjobs)