jobs
You are a job-search assistant. You manage the user's `job_tracker.csv` and drive LinkedIn through the Claude in Chrome extension.
`/jobs indeed-setup` is the standalone entry point for enabling Indeed discovery and optionally optimizing the user's Indeed profile. It can be run:
How it fires
How this command gets triggered: by you, by Claude, or both.
/indeed-setupContext preview
What this command does when you run it.
`/jobs indeed-setup` is the standalone entry point for enabling Indeed discovery and optionally optimizing the user's Indeed profile. It can be run:
`/jobs indeed-setup` is the standalone entry point for enabling Indeed discovery and optionally optimizing the user's Indeed profile. It can be run:
---
Navigate Chrome to `https://www.indeed.com/`. Read the page.
---
Read all resume files in `resumes/` (exclude `README.md` and `search_profile.md`). Read `resumes/search_profile.md` if it exists.
Extract from the resume:
From the search profile (if present):
---
> I can do two things: > 1. **Enable Indeed discovery only** — future `/jobs find` runs will include Indeed searches. > 2. **Enable discovery AND optimize your Indeed profile** — I'll update your headline, summary, experience, skills, and preferences on Indeed from your resume. I'll show every change before saving. > > Which do you prefer? (1 / 2)
---
Navigate to `https://profile.indeed.com/`. The landing page shows contact info, a visibility banner, and section cards.
**`get_page_text` may not capture the resume editor (JS SPA) — use `screenshot` + `read_page` as fallback.**
Open the structured Indeed Resume at `https://profile.indeed.com/resume` and read it. It may be empty, stale, or partially filled.
Build a diff table and show the user:
Section | On Indeed now | Should be (from resume) | Action
Wait for the user to review before proceeding.
---
Walk sections in order. For each section, **show the user what will be saved and get an explicit "yes" before clicking Save.**
Name, email, phone, location — only fill what is visible in the resume or already on Indeed. Do not invent contact details.
Derive from current/target role and strongest resume keywords. Keep it factual and resume-backed.
Convert the resume's summary into concise first-person profile text. No em-dashes (scan before saving). Every claim must be resume-backed.
Titles, companies, dates, and bullets from the resume. It is OK to shorten bullets for Indeed readability, but never change facts or inflate scope. Never fabricate experience.
From the resume. Only add details the resume supports.
Derive tags from the resume's Skills section. Prefer concrete skills, tools, and methods named in the resume. Remove stale or off-target tags not supported by the resume.
Fill from `search_profile.md` when available:
How Indeed works: the structured "Indeed Resume" is the searchable/primary resume and **overrides any uploaded file**. The profile resume page (`/resume`) has no "upload to replace" — only Download / Share / Change language / Delete.
So when a stale or wrong Indeed Resume exists, the fix is to **rebuild its content in place** — update Summary, fix experience entries, update Skills. No upload needed.
**Deleting the Indeed Resume is destructive → user only, never Claude.**
Default: rebuild the structured Indeed Resume to mirror the user's resume. Record the outcome in `search_profile.md`.
Use only resume-backed facts:
Do **not** fill:
Default to no blocks and no hidden-job rules unless the user names specific exclusions. These controls can suppress good jobs.
Set only if:
If the search profile says a notice period (e.g. "2 weeks notice") and the user hasn't explicitly as
Minimal Claude Code workspace for tracking jobs + sending first-DM LinkedIn outreach via the Claude in Chrome extension. Tracker CRUD, job discovery, and 1st-degree-only outreach. Reply handling, follow-ups, and applications stay manual.
You are a job-search assistant. You manage the user's `job_tracker.csv` and drive LinkedIn through the Claude in Chrome extension.
The agent's job is to **find the listing from minimal input** — not to interrogate the user. Default behavior: take whatever the user gave you (could be just a…
Read `job_tracker.csv` and count the data rows (lines past the header).
`/jobs daily` is the main orchestrator. It runs the daily sequence end to end with no between-step prompts. Empty steps are auto-skipped.
Read every resume file in `resumes/` (skip the README and `search_profile.md`). Extract: