AI Workflows  |  Agentic Tooling  |  Operations  |  Process Documentation

---

Job searching is an operations problem. I’ve sent over a hundred applications. I know what the process actually looks like from the inside — the tab switching, the copy-pasting, the mental overhead of trying to decide whether a posting is worth your time before you’ve even read it properly. It adds up fast and burns you out faster.

So I built a workflow that handles the search layer for me. Not a script. Not a browser extension that scrapes and dumps a CSV. I used Claude for Chrome — Anthropic’s browser AI — to open tabs, navigate live job boards, read actual postings, and score them against my real profile. All in one session.

This documents exactly how I built it and how it works.

---

What Claude for Chrome Actually Does

Claude for Chrome is a browser extension that opens a sidebar inside your active Chrome window. It can see what’s on your screen, navigate to URLs, open new tabs, click through pages, and read live content — all while you watch it work and stay in control. It’s not running in the background somewhere. It’s operating in your actual browser session, which means it uses your existing logins, your cookies, your authenticated state.

That matters a lot for job searching, because most job boards require an account to see full listings. Claude for Chrome doesn’t need to crack that — it just uses the session you already have open.

---

Step 1: Ground the System With Your Real Profile

Before Claude touches a single job board, you give it context. This is the most important step. The quality of the scoring depends entirely on what you feed it upfront.

I uploaded three things into the sidebar:

• My resume as a clean text or .md file
• My LinkedIn profile saved as a PDF
• A short metadata block listing my target job titles, preferred salary range, location requirements (remote, hybrid, or on-site), and any hard deal-breakers

That last item is critical. If you don’t define your deal-breakers explicitly — “must be fully remote,” “no roles below $X,” “no staffing agencies” — Claude will surface everything and you’re back to manual filtering. The constraints are what give the scoring engine its teeth.

---

Step 2: Log Into Your Job Boards First

This is a one-time setup and it’s simple: before you run the workflow, manually open tabs for the platforms you want to search and make sure you’re logged in. Claude for Chrome operates inside your active session — it’s not logging in for you from scratch, it’s piggy-backing on the authentication you already have.

The first time you run this, you’ll need to do this manually for each platform. After that, your browser handles it automatically on return visits as long as your sessions haven’t expired.

The platforms I used for my search:

• LinkedIn
• Indeed
• ZipRecruiter
• Glassdoor
• Built In
• Remotive
• FlexJobs
• Teal HQ
• Remote Rocketship
• Talentify
• Wellfound
• Himalayas.app
• Remote.co
• Google Jobs (via search)

You don’t have to use all of these. Narrow the list to whatever matches your field and your target work arrangement.

---

Step 3: Run the Workflow — And Approve Before It Moves

This is where it gets interesting. Once your profile is loaded and your sessions are live, you give Claude a structured prompt telling it what to do. Claude doesn’t just run — it summarizes the planned steps first and asks you to approve or edit before it starts.

That approval step matters. It’s not just a safety check — it’s where you can refine the scope, adjust the scoring weights, or add a platform you forgot. Once you confirm, Claude opens tabs automatically, navigates each job board, reads through listings, runs the comparison against your profile, and builds the output as it goes.

You watch it happen in real time.

---

The Prompt I Used

Drop this into the Claude for Chrome sidebar after uploading your files. Edit the bracketed sections for your situation:

I have uploaded my resume, LinkedIn profile, and a list of my job search criteria including target titles, salary range, location preferences, and deal-breakers. Use those files as my baseline profile.

I have active browser sessions open for the following job boards:
[LIST YOUR PLATFORMS HERE]

Please do the following:

1. Navigate to each platform and search for: [YOUR TARGET JOB TITLES]
2. Read the top listings on each platform — aim for 8 to 10 per board.
3. Score each listing against my profile using this weighted breakdown:
   - Skills and experience alignment: 40%
   - Scope of role and operational responsibilities: 30%
   - Work arrangement and location match: 30%
4. Exclude any listings that match these deal-breakers: [YOUR DEAL-BREAKERS]
5. Output the results as a clean table with these columns:
   Job Title | Company | Match % | Key Reason | Direct URL

Before you start, summarize the plan and wait for my approval.

---

Step 4: Get the Results and Save Them

When the workflow finishes, Claude delivers a structured table in the sidebar — job title, company, match percentage, the primary reason for that score, and a live link to each posting.

To save it: copy the table from the sidebar and paste it directly into a Google Doc. That gives you a clean, shareable record you can track over time, annotate, and return to. You can also paste it into a Google Sheet if you want to sort or filter by score.

This is the simplest, most reliable export method. No conversion needed — Claude’s markdown table pastes cleanly into both formats.

---

A Few Things Worth Knowing Before You Run This

• Session timeouts: If a platform redirects Claude to a login page mid-session, it means your authentication expired. Just manually log back in on that tab and let Claude continue.
• Be specific with your titles: Vague search terms return vague results. The more precise your target titles, the better the signal-to-noise ratio in the output.
• The scoring is only as good as your inputs: A weak profile document or a missing deal-breaker list produces generic scoring. Take ten minutes on the setup and the output pays you back.
• This scales: The same workflow works for any research task that involves reading across multiple websites and scoring against a fixed set of criteria — competitive research, vendor evaluation, content audits. The job search use case is just where I built it first.

---

Why This Works

The job search problem isn’t a lack of listings. It’s attention cost. Every time you manually open a posting, read the full description, and decide it’s not worth applying to, that’s a small but real drain. Multiply that by a hundred applications and you understand why the process is exhausting even before you get to the actual work of applying.

This workflow moves the filtering upstream. You define your criteria once, Claude does the reading and scoring, and you receive a prioritized list. You spend your attention on the top matches instead of burning it on the noise.

That’s the whole idea.