# Job scout template for Claude Code

> Source: <https://gist.github.com/catherine-grabar/8ba71c1e96c527e541d04223cc5ec909>
> Published: 2026-11-02 22:26:54+00:00

A weekly job search that checks company career pages for you, confirms every posting is still live, and ranks what it finds against your own criteria.

Made by Catherine Grabar. Free to use and change.

1. Make a new folder on your computer, for example `job-search` .
2. Save this file in it.
3. Open the folder in Claude Code and type:
**"Read job-scout-template.md and set up my job scout."**

Claude will interview you, create the files below, and run your first search. After that, run it
whenever you want a fresh sweep by typing **"Run my job scout."** You can also ask Claude to put it
on a weekly schedule.

**The most important step is the brain dump.** The scout is only as good as what it knows about
you. Ten minutes of rambling about your past projects will do more for your results than any
search setting.

You are setting up and running a job scout for the person in this folder. Speak plainly and define any technical term the first time you use it; they may be new to Claude Code.

Create these files by interviewing the person. Ask a few questions at a time, not all at once. Encourage them to answer by dictation or in messy notes; you clean it up.

`about-me.md` (the brain dump)

- Every role they've held: company, title, dates, what the product was, who used it.
- Three to five projects they're proudest of: the problem, what they did, what changed.
- Skills and tools, and which ones they'd want to use every day.
- Unusual background that sets them apart (an industry, a type of user, a side project).
- What they're worst at or don't want to do anymore. Ask follow-up questions until each project has a concrete outcome. Vague answers produce vague rankings.

`what-im-looking-for.md` (the criteria)

- Role titles and levels to search for, plus titles to ignore.
- Industries or missions they care about, ranked.
- Location: remote, specific cities, willingness to relocate, countries.
- Company size and stage.
- Pay: their floor and what they'd count (base only, or total with equity).
- Hard no's in their own words (for example "no gambling apps," "no teams over 10 people").
- What makes a role a top pick for them.

`companies.md` (the watchlist)

- Companies they'd love to work at, plus any they want to avoid.
- For each company, find which job board it uses and record the board name and the company's ID on that board (see "Where to look" below). This list grows every run.

`applied.md` (so the scout skips them)

- A simple table: company, role, date applied, status. Ask them to paste in what they've already applied to. Keep it updated whenever they mention a new application.

`resume.md` (optional, for the resume check)

- Ask them to paste in the text of the resume they're currently sending.

Show them each file when it's done and ask what's wrong or missing before moving on.

**Read first, search second.** Before any search, read all the files above and the most recent
`scout-YYYY-MM-DD.md` in this folder, if one exists.

**Where to look.** Run both halves; either one alone misses roles.

1. *Watchlist check.* Most companies post jobs through one of a few public job board systems,
and several have free public feeds you can read directly:
  - Greenhouse: `https://boards-api.greenhouse.io/v1/boards/<company-id>/jobs`
  - Ashby: `https://api.ashbyhq.com/posting-api/job-board/<company-id>?includeCompensation=true`
  - Lever: `https://api.lever.co/v0/postings/<company-id>?mode=json`
  - Others to check by hand or by their own feeds: Workday, Rippling, SmartRecruiters, Workable,
BambooHR.
Check every company in `companies.md` . If you don't know a company's ID, try its name in
lowercase with no spaces, then look at the careers page to confirm.
2. Greenhouse: 
3. *Discovery.* A watchlist only finds companies the person already knows about. Also search
broadly: general job sites, startup directories (the Y Combinator directory has a public data
mirror at[https://yc-oss.github.io/api/](https://yc-oss.github.io/api/) ), and industry news about recent fundraising. Treat
every result from a general job site as a lead only, never as a confirmed listing.

**Verify every link.** Open each posting on the company's own careers page or job board and
confirm it loads with the stated title today. Many listings on general job sites have already
closed. Two traps: some boards return a normal page that says "job not found" for a closed
posting, so check the content, not just whether the page loads; and some large-company career
sites wrongly say a live job is closed, so flag those for a hand check instead of dropping them.

**Cross-check `applied.md`.** Never recommend a role they've already applied to. If they have an
application in progress at a company, flag any new role there with the status of the existing one.

**Rank by fit, not by brand.** Use `what-im-looking-for.md` and `about-me.md`. For each role, say
in one or two sentences why it fits this person specifically, citing their real experience. Be
honest about weak fits and caveats (onsite five days a week, a level stretch, pay not listed).
Never invent pay numbers; "not listed" is a fine answer. When two roles fit about equally, rank the
newer posting higher, since applying early matters.

**Check the resume against each top pick** (if `resume.md` exists). Read the resume cold, the way a
stranger or an automated screener would, and compare it with the job's requirements. Label each top
pick:

- *Ready:* the resume already shows what the job asks for.
- *Resume gap:* they have the skill or experience (it's in`about-me.md` ) but the resume doesn't show
it. Name the requirement and suggest where it could go.
- *Real gap:* they don't have it. Say so plainly so they can decide whether to apply anyway.
Never suggest adding anything to the resume that`about-me.md` doesn't support.

**Write the results** to `scout-YYYY-MM-DD.md` with today's date:

1. A short header: what you checked, today's date, any rules applied.
2. Top picks, about 10, numbered by fit. Each one: company, title, location, pay if listed, posting date if shown, why it fits, caveats, the resume label, and the direct link.
3. Other live roles worth a look, in a short table.
4. Skipped because already applied.
5. Notes on the market: what's quiet, what's hiring, how many links were dead.
6. Every company you checked that had nothing relevant. This makes the next run faster; don't skip it.

Add any new companies and their job board IDs to `companies.md` so the next run checks them.

Then give the person a short summary in chat: the top three to five picks with a one-line reason each, plus anything time sensitive.

**Learn as you go.** When the person disagrees with a ranking, rejects a type of role, or finds a
good role on their own that the scout missed, ask whether to add a rule to
`what-im-looking-for.md` or a source to `companies.md`, and record their answer. The scout should
get better every week.

- LinkedIn and some large companies block automated searches. Check those by hand.
- No search finds everything. A role they find on their own is a signal to add a new source.
- Postings close fast. Treat a scout as good for a few days.
