A daily job search agent that finds more than listings
The demo shows two things most people don't do: automated daily scanning for relevant roles, and signal detection in your LinkedIn feed that catches hiring intent before anyone posts a job. Here's how to build both.
Sam Tasman-Jones·~12 min read·Claude Projects + Zapier or Make
The demo has three views: a daily job listing feed (structured summaries with match reasoning, not raw listings), a signals feed (hiring intent spotted in your LinkedIn timeline before a job is posted), and a target companies tracker where you keep your list warm and draft outreach.
The key insight is that most job searches are passive — you check boards when you feel like it, and you only see what companies choose to advertise. This flips that. The agent runs on a schedule, the feed scanner looks for signals that most people scroll past, and the companies list makes sure you're never starting cold.
The listing scan is straightforward to replicate with a Claude automation. The LinkedIn signal detection is more sophisticated — it works by taking a screenshot of your feed and having Claude parse it for indirect hiring signals. Both are described below.
What you need
🤖
Claude.ai Pro accountFor Projects with persistent context. Free tier works for one-off prompts but not scheduled runs.
⚡
Zapier, Make, or n8n (optional)For scheduling the daily scan and delivering results to Slack or email. Not needed for the manual version.
📋
Your target role criteriaA clear brief: role titles you want, industries, seniority level, location constraints, what you're not interested in. The sharper this is, the better the match scoring.
📸
A way to screenshot LinkedIn (for signals)The cowork routine works by screenshotting your feed. You can do this manually or use a browser automation tool.
⚡ Option 1 — No code
Run a job search briefing manually
This uses Claude.ai to build your search brief, find and score listings, and surface signals — all inside one chat. No automation, no manual copy-pasting of job boards. Best run each morning as a 10-minute ritual.
1
Let Claude interview you on your preferences first
Don't try to write the perfect role brief yourself — start a new Claude.ai chat and run this prompt. Claude will ask you the right questions and build the brief for you:
Interview prompt
I want to set up a job search agent that finds and scores roles for me daily. Before we build the search brief, I need you to interview me on my preferences so the brief is accurate and specific to me.
Ask me one question at a time. Cover: the role types and titles I'm targeting, industries and company types I prefer, seniority and stage, location constraints, what I'm actively not interested in, and anything else that would help score a job listing for fit.
Once you've asked everything you need, generate a complete job search brief I can use as a Claude Project instruction — formatted and ready to paste in.
💡
This takes 3–5 minutes but means the brief actually reflects how you think about fit — not a generic template. Claude will ask things you wouldn't have thought to specify.
2
Paste the brief into a Claude Project and run the daily search
Create a Project called "Job Search". Paste the brief Claude generated in step 1 into the Project instructions. Then each morning, open a new chat in the Project and let Claude do the searching:
Daily search prompt
Search for new job listings posted in the last 24 hours that match my brief. Check Seek, LinkedIn, and Wellfound. For each role that scores 70+, give me:
• Match score: X/100
• Why it matched: 3 specific bullets tied to my criteria
• Red flags: anything that doesn't fit
• Draft outreach: a short, warm LinkedIn note I could send to the hiring manager or a relevant connection
Surface the top 3–5 only. Skip anything below 70.
💡
Claude can search job boards directly in claude.ai — you don't need to find and paste listings yourself. If it can't access a board directly, it'll tell you and you can paste the URL or a few listings to supplement.
3
Optional: scan LinkedIn for informal hiring signals
This step is for people using Claude Cowork or a browser automation tool that can scroll and screenshot your LinkedIn feed — it's not a manual copy-paste step. The idea: your feed is full of indirect signals (founders posting about scaling, execs talking about product gaps) that appear weeks before a job is posted. If you have a tool that can screenshot your feed automatically, upload the images with this prompt:
Feed signal prompt
Here are screenshots of my LinkedIn feed from today. Scan every visible post and identify:
1. Hiring signals — headcount, team growth, new hires, open roles, "we're scaling"
2. Company growth signals — funding rounds, new products, market expansion
3. Indirect signals — founders or execs posting about product challenges that suggest PM headcount is coming
For each signal: source name, signal type, 1-sentence relevance note against my role brief.
💡
Claude Cowork (or similar tools) can automate the scrolling and screenshotting — you just get the signal digest delivered. Without automation, this step isn't worth doing manually every day.
⚙️ Option 2 — Start building
Make it run on a schedule
The automated version runs every day at 5pm, searches job boards, scores matches against your brief, and drops a structured briefing into Slack. Two ways to set this up: a Claude Code agent (simplest if you're comfortable with a terminal) or a Make/Zapier workflow (no code required, ~2 hours).
1
Get your system prompt
Whichever method you use, this is the core prompt that drives the agent. Customise the scoring criteria to match your brief from Option 1:
Agent system prompt
You are a daily job search agent. Search Seek, Wellfound, and LinkedIn Jobs for new listings posted in the last 24 hours matching the following brief:
[Paste your role brief here — output from the Step 1 interview]
For each listing that scores 65+, output:
{
"company": "Company name",
"role": "Role title",
"location": "Location",
"match_score": 0-100,
"why_matched": ["bullet 1", "bullet 2", "bullet 3"],
"red_flags": ["any red flags or empty array"],
"outreach": "Short LinkedIn message draft — warm, specific, not generic",
"source": "URL or board name"
}
Return a JSON array of matched roles only. Skip anything below 65.
2
Route A — Claude Code agent (recommended if you're technical)
Claude Code can run as a scheduled agent directly from your terminal — no third-party automation tools needed. Set it up with a cron job or use Claude Code's built-in scheduling. Use this as your starting prompt in Claude Code:
Claude Code build prompt
Build me a job search agent that runs daily at 5pm. It should:
1. Search Seek (via RSS), Wellfound (via API or scrape), and optionally LinkedIn Jobs for listings posted in the last 24 hours
2. Send each listing to the Anthropic API (claude-3-5-haiku for cost efficiency) with my scoring system prompt
3. Filter to matches >= 65 and format them as a readable Slack message (one block per role: company, title, score, why it matched, outreach draft)
4. Post to my Slack channel via webhook
Use a cron schedule (5pm daily). Store my role brief in a .env file so I can update it without touching the code. Log each run to a local JSON file for history.
Keep it simple — single Python script, minimal dependencies.
💡
Seek has a public RSS feed per search query (search Seek, copy the URL, add /rss). Wellfound has a public API. LinkedIn doesn't — use Otta or Greenhouse job boards as supplements.
3
Route B — Make or Zapier (no code)
In Make or Zapier: schedule trigger (daily 5pm) → HTTP module to fetch Seek RSS → Claude AI module with your system prompt → parse JSON response → Slack message per matched role. No terminal required — get a developer to set it up in an afternoon if needed.
💡
Make has a native Claude module. In Zapier you'll use the HTTP action to call the Anthropic API directly. Both work — Make is slightly cleaner for multi-step JSON parsing.
4
Add the target companies tracker
The companies tracker is a separate Claude Project where you store your shortlist. Add a company, Claude searches for recent signals (funding, headcount posts, product launches) and logs them alongside your warm/cold status:
Company tracking prompt
I want to track [Company name] in my target companies list. Here's what I know:
— Why I'm interested: [Your reason]
— Role I'd want: [Role title]
— Connections I have there: [Names if any]
Search for any recent signals — LinkedIn posts, funding news, product launches, headcount changes. Log this company as [Warm/Cold/Watching] and suggest a first-contact message for my most relevant connection.
Going further
→
Add a "track this company" Slack command — type /track [company] and the bot logs it to your companies list and schedules a weekly signal check
→
Connect to a Notion database to track applications — each matched role automatically creates a card with status, outreach draft, and signal log
→
Add a second Claude pass on your LinkedIn connections to surface who you know at each matched company — prompt: "Who from my connections list works at X?"
Want to build something like this for your team?
I build AI-powered tools for product, sales, and ops teams — prototypes to production-ready. Let's talk.