job-tracker-setup

job-tracker-setup

A Claude skill that builds a live, self-updating job-application tracker from your Gmail. Point it at your inbox and it scans for application confirmations, rejections (including deleted ones in Trash), interview invites, offers, referrals, and recruiter/HM correspondence, then publishes a live dashboard that stays current on a schedule you choose.

Created by Jessica Higgs. Generalized from a workflow originally built for one real job search, so it can bootstrap a fresh tracker for anyone's inbox.

What it looks like

A live dashboard, updated automatically from your inbox — every application with its status, date, and notes, plus quick filters and search across companies and roles. (Sample below uses illustrative data.)

What it does

• Seeds from an existing spreadsheet/CSV (optional) or builds your pipeline from scratch by scanning inbox history.
• Classifies each application into a small, ordered status set: Applied → Warm → Interviewing → Offer, with Rejected as a terminal state.
  • Warm = the specific role didn't proceed, but a recruiter is keeping you warm / referring you elsewhere — a state that otherwise gets mislabeled.
• Maintains the tracker on every run: adds new applications, moves statuses forward on real email evidence, and appends dated notes from recruiter threads — without ever overwriting your existing notes.
• Publishes a searchable, filterable live dashboard (see job-tracker-setup/assets/tracker_template.html) that you can reopen anytime.
• Optionally schedules itself to re-sync automatically (e.g. nightly).

Requirements

• A connected Gmail MCP connector (to search/read mail).
• Cowork's artifact tool (mcp__cowork__create_artifact / update_artifact) to publish the live dashboard.
• Optional: a Drive/Sheets connector (only to seed from an existing spreadsheet) and the schedule skill (only for automatic recurring runs).

Install

This is a Claude skill. You install it once, then talk to Claude normally — you never open or read the files yourself. Claude reads them for you.

1. Download the skill

On this page: click the green Code button → Download ZIP, then unzip it.

2. Zip the inner folder

⚠️ This is the step people get wrong. The download gives you two nested folders with almost the same name:

job-tracker-setup-main/        ← the repo. NOT this one.
└── job-tracker-setup/         ← the skill. THIS one.
    ├── SKILL.md
    ├── references/
    └── assets/

You want the inner job-tracker-setup folder — the one with SKILL.md sitting directly inside it. Right-click it → Compress.

If you zip the outer folder by mistake, SKILL.md ends up one level too deep and the upload will be rejected.

3. Upload it to Claude

Go to claude.ai → Settings → Capabilities → Skills → upload your zip.

4. Connect Gmail

Settings → Connectors → add Gmail, and sign in with the address you apply to jobs from.

The tracker only ever sees this one account. If you apply from a second address, those applications won't show up.

5. Ask for your tracker

Start a new chat and say:

build me a job application tracker from my Gmail

Claude will ask you a few short questions — which account to scan, whether you have an existing spreadsheet to import, whether you use a Gmail label for job mail, and whether you want it to re-check automatically each night. Defaults are fine for all of them; answer "no" to anything you don't have.

Then it scans your inbox and builds your dashboard.

Later

To refresh it any time:

sync my job tracker

Using Claude Code instead? Drop the inner job-tracker-setup folder into ~/.claude/skills/ and the skill will load. Note that the live dashboard is published as a Claude artifact, so the final publishing step is designed for claude.ai.

How it works (files)

• SKILL.md — the entry point: prerequisite checks, the setup interview, and the build/schedule flow.
• references/gmail_queries.md — the exact Gmail search patterns for each signal type (new application, rejection, interview, offer, sent-mail dialogue, referral) and the reasoning behind them.
• references/status_rules.md — the status/notes rules that keep the tracker trustworthy across many runs (progression, "Warm is sticky," re-applications, row matching).
• assets/tracker_template.html — the self-contained dashboard template with {{LAST_SYNC}}, {{RECENT_CHANGES}}, and {{DATA}} placeholders.

Design notes / hard-won lessons

This skill has been stress-tested against a real, active job search. A few rules exist specifically because the naive version got them wrong:

• Never trust a single phrase to detect an application. Confirmation wording varies enormously; the skill matches on a wide phrase list and a sender heuristic and a direct label sweep.
• Read every email's body for the exact role. One email thread can contain confirmations for two different roles at the same company. Companies don't send duplicate confirmations, so two confirmations with different role text are two applications — never merge them.
• Convert Gmail's UTC timestamps to the user's local date before deciding what counts as "today."
• A tracker only sees the one account it's connected to — mail from a second address won't appear.

Limitations

• Single Gmail account at a time.
• English-language emails — the bundled search phrases are English; other languages need translated patterns.
• Validated end-to-end against one real inbox — a very differently-shaped mailbox may surface untested edge cases.

License

MIT — see LICENSE.