I Built an AI Skill That Searches Thousands of Career Pages
I Built an AI Skill That Searches Thousands of Career Pages
Job boards only show you the jobs that companies pay to post there. Plenty of great roles live only on a company's own careers page, and nobody has time to check hundreds of those by hand. A friend recently shared someone's LinkedIn post with a prompt for using AI to search job boards.
I'm a fan of this kind of help from AI, but I thought the prompt could go a bit further.
So I built a skill that takes it further. Ask it for new jobs, and it searches thousands of career pages, checks that each posting is still open, scores how well each one fits the person searching, and hands back a ranked list with links. Here's how it works and how you can build your own.
What a skill is
A skill is a folder of instructions that an AI reads whenever a task calls for it. Think of it as a playbook: instead of explaining your process every time, you write it down once, and the AI follows it.
It has three files:
job-finder/
├── SKILL.md ← the step-by-step instructions
└── references/
├── profile.md ← who I am and what I want
└── search-patterns.md ← where and how to search
The top of SKILL.md has a short description telling the AI when to use it. The description is broad on purpose: any time someone asks about new jobs, open roles, or what's hiring, the skill kicks in, even if they never mention it by name.
Step 1: Write a profile that's specific
The profile is what turns "here are some jobs" into "here are jobs for you." A resume alone isn't enough, because it says what you've done, not what you want next. A good profile covers:
- Target titles: primary roles, adjacent roles you'd still consider, and roles you don't want
- Skills by level: expert, strong, learning, and what you don't have, so the skill can flag honest gaps
- Location and work style: remote, hybrid, or on-site, plus time zone
- Pay: a hard floor and an ideal range
- Target companies: a short list of employers to always check
- Target industries and company size: so it can find employers you haven't heard of
- Dealbreakers: industries, contract terms, or office requirements that rule a job out
The dealbreakers list does more work than you'd expect. Without it, you get a pile of technically matching jobs you'd never take.
Step 2: Search in three layers
You can't search "every career page" directly. But you can get surprisingly close by searching in three layers.
Layer 1: The platforms behind career pages
Most companies don't build their own hiring system. Their careers page runs on a platform like Greenhouse, Lever, Ashby, or Workday, and every job on those platforms lives at a predictable web address. That means one search can reach thousands of employers at once:
site:boards.greenhouse.io "analytics engineer" remote
site:jobs.lever.co "data analyst" dbt
site:myworkdayjobs.com "senior data analyst" Denver
The skill covers 14 of these platforms, including SmartRecruiters, Workable, iCIMS, BambooHR, and Rippling.
Layer 2: Companies you already want
For every company on the target list, the skill searches for its careers page by name and then searches inside the company's own website. Some big employers host jobs on their own domain, and this layer catches them.
Layer 3: Companies you haven't found yet
This is my favorite part. The skill searches for lists of top companies in the target industries, plus recent startup funding news, since a company that just raised money is usually about to hire. It also uses search tricks like inurl:careers to find career pages directly. Then it checks those new companies the same way as Layer 2 and suggests any good ones to add to the target list.
Step 3: Check every posting, then score it
Search results are full of stale listings, so the skill opens each posting before it counts. It drops anything that's closed, older than 30 days, posted by a recruiter with no named employer, or hits one of the profile's dealbreakers. The same job often shows up on both a hiring platform and the company's site, so it removes duplicates too.
Then it scores each job out of 10:
| Category | Points | What it checks |
|---|---|---|
| Skills match | 4 | How many of the must-have requirements the profile covers |
| Seniority | 2 | Whether the title and years of experience line up |
| Logistics | 2 | Location, remote policy, and pay if it's listed |
| Preference bonus | 2 | Target company, target industry, or nice-to-haves |
Only jobs scoring 6 or higher make the list. For the top five, it writes a sentence on why the job fits and one honest gap. That gap is useful: it shows exactly what to address in a cover letter.
How to build your own
You don't need to write code. Here's the fastest path:
- Ask an AI like Claude to help you create a job-finder skill. It can draft all three files for you.
- Replace the sample profile with your own details. Be blunt about dealbreakers and your pay floor.
- Install the skill in your AI assistant under Settings, in the Skills section.
- Start a new chat and ask, "Find me new jobs I'd be a good fit for."
- Review the results, then tweak the profile. If it keeps surfacing roles you don't want, add a dealbreaker.
What it can't do
It's worth being honest about the limits. It can't see jobs behind a login, like some LinkedIn listings or internal job boards. It runs when you ask, not automatically every morning. And it doesn't remember which jobs you've seen between chats, so the profile ends with a short "already applied" list to filter those out.
It also won't apply for you. That part is still on you, and honestly, I think it should be.
The real win
The biggest benefit isn't speed. It's that a job seeker is no longer limited to the jobs that happen to be advertised where they're looking. The skill finds roles at companies they'd never have thought to check, and it explains why each one fits.
Job hunting is still hard. But spending energy on applications instead of endless scrolling makes it a lot more bearable. If you're looking, or you know someone who is, give it a try. And if you build your own version, I'd love to hear what you change.