- Home
- Skills
- Workflow Automation
- Job Source Fetcher - ATS + Career Page Job Scraper
Works with the AI Agents you already use
Job Source Fetcher - ATS + Career Page Job Scraper
Pull openings straight from public ATS APIs and robots-gated career pages, and never confuse a dead scraper for a quiet job market.
$12.99
Job Source Fetcher - ATS + Career Page Job Scraper
Example session with this skill installed
Check Figma, Ramp and Zerodha for platform and SRE openings.
- Read your context and instructions
- Compiled the job source fetcher
Config: Figma on Greenhouse, Ramp on Ashby, Zerodha with only a careers page.
$ python3 skills/job-ats-fetch/scripts/fetch_ats.py --out ats.json fetching 2 board(s), 1.0s between requests, as 'jobdigest/1.0 (+you@example.com)' Figma greenhouse boards-api.greenhouse.io Ramp ashby api.ashbyhq.com fetch report: Figma ok 5 of 145 posting(s) matched (134 off-role, 0 off-location), stopped at the fetch.max_per_company cap of 5 after 139 of 145 Ramp no-match 0 of 171 posting(s) matched (171 off-role, 0 off-location) total: 5 job(s) from 2 board(s); 0 errored wrote ats.json $ python3 skills/job-careerpage-scan/scripts/scan_career_pages.py --out pages.json scanning 1 page(s), 1.0s between requests, as 'jobdigest/1.0 (+you@example.com)' Zerodha no-match 0 link(s) matched a configured role out of 0 anchor(s) total: 0 job(s) from 1 page(s); 0 errored, 0 skipped by robots.txt wrote pages.jsonThree companies, three genuinely different outcomes, and you can tell them apart at a glance.
Figma:
ok, with its own arithmetic shown. Five kept out of 145 published. 134 were off-role.
The cap stopped the scan on the 139th, so six postings were never examined — and the line says so,
rather than implying all 145 were considered and rejected.Ramp:
no-match, not an error. The board answered, 171 postings exist, none matched these
role keywords. Your filters are narrow; nothing is broken. Had the slug been wrong instead, the
line would readerror HTTP 404 - is the board slug right?and the run would have exited
non-zero.Zerodha:
0 anchors, which is the honest answer. Their careers page renders its job list in
JavaScript, so there is nothing in the HTML for any scanner to read. The report says zero
anchors, not zero jobs — the distinction between "I looked and found nothing" and "there was
nothing to look at." A headless browser would get those jobs; this pack does not ship one, and
reference/limits.mdsays so plainly.And a company whose
robots.txtdisallows the path never gets fetched at all:Beta Corp skipped robots.txt disallows this path for 'jobdigest'Zero page requests for that company. The
robots.txtis fetched first, every time, and that is
not a setting worth turning off.Then hand it to the scheduler:
$ python3 skills/job-digest-schedule/scripts/render_schedule.py --at "sunday 09:00" --out .github/workflows/job-digest.yml
Asia/KolkataSunday 09:00 becomescron: "30 3 * * 0". The rendered workflow has least-privilege
permissions,persist-credentials: false, every action pinned to a major version, a
timeout-minutes, a concurrency group that does not cancel a run in progress, and the SMTP
password scoped to the one step that sends. It contains no|| true, no2>/dev/nulland no
continue-on-error, because a scheduler that hides a failure is the whole problem this pack exists
to solve.
Connects securely to your tools. The creator never sees your data.
About this skill
Every job scraper eventually lies to you. A board slug changes, a page starts rendering in JavaScript, a host times out — and the pipeline reports zero new jobs, which looks exactly like a week in which nobody was hiring. You find out six weeks later.
This pack is built around refusing to do that. Every company gets a line in a fetch report saying what happened to it, the report travels inside the output JSON rather than only in the logs, and more than half the sources failing exits non-zero rather than writing an empty list.
Three skills
Skill What it does job-ats-fetchReads openings from five public ATS JSON APIs — Greenhouse, Lever, Ashby, Workable, SmartRecruiters — filtered by your role keywords and locations. No API keys, no accounts, no HTML parsing. job-careerpage-scanFor companies with no supported ATS: a robots.txt-gated anchor scanner that fetches robots.txtbefore the page and skips the company if it is disallowed.job-digest-scheduleRenders a scheduler from your config — GitHub Actions YAML, a crontab line, or a Windows Task Scheduler XML — with your local send time converted to UTC for you. Working in two minutes
python3 skills/job-ats-fetch/scripts/fetch_ats.py --list-providers python3 skills/job-ats-fetch/scripts/fetch_ats.py --provider greenhouse --board figma --no-filterThat second command needs no config at all. Once you have a
jobdigest.config.jsonwith acompanies[]list,fetch_ats.py --out ats.jsondoes the whole set.Made to be customized
One file,
jobdigest.config.json, in the directory you run from.
You change It affects companies[].ats{"provider": "greenhouse", "board": "figma"}— five providers, andreference/finding-your-board-slug.mdis a three-step method for finding any company's slug.companies[].careers_urlUsed only when a company has no ats. Scanned subject to robots.txt.roles[],search.locationsWhich postings are kept. Keywords match the title as whole words, so sredoes not fire inside "pressure".fetch.request_delay_s,timeout_s,max_per_companyPacing and ceilings. fetch.user_agentSent on every request. Put a real contact address in it. fetch.respect_robotsLeave it true. Setting itfalseprints a loud warning and is your own decision and liability.profile.timezone, and the schedule you passWhat job-digest-schedulerenders.Etiquette is enforced, not just documented
An honest
jobdigest/1.0 (+your@email)User-Agent on every request, naming your tool and a way to reach you. Afetch.user_agentcontainingMozillais rejected at the socket — the pack will not pretend to be a browser even if you ask it to.robots.txtis fetched once per host and parsed to RFC 9309 (longest match wins,Disallowwins a tie); an unreachablerobots.txtis treated as disallowed, because unconfirmed is not permission. A disallowed path costs zero page requests.The output is a contract, not a format
Both fetchers emit the same canonical JSON —
title,company,url,location,date,_portal— documented field by field inshared/job-schema.md. It feeds the companion Job Digest Reporter pack unchanged, and it is plain enough to feed your own spreadsheet, your own database, orjq.Compatibility
Environment Support Claude Code · claude.ai · Claude Agent SDK Skills auto-discovered from skills/Any other coding agent (Cursor, Windsurf, Copilot, Codex, Gemini CLI, …) Works — point the agent at USAGE.mdor the relevantSKILL.mdPlain terminal, no agent at all Works — USAGE.mdis copy-pastePython 3.10+, standard library only —
urllibandhtml.parser, nothing to install. macOS, Linux, Windows. No Node, no Bun, no headless browser, no API keys, no accounts, no telemetry. The only hosts contacted are the ones in your own config, and every one is printed before the request.Includes MIT-licensed portions of ai-job-search (© 2026 Mads Lorentzen); modifications and new work © 2026 Sujal Goswami. Full notice:
shared/LICENSE-MIT.md. Not affiliated with or endorsed by ai-job-search.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Workflow Automation.
- Five ATS providers, not all of them. Greenhouse, Lever, Ashby, Workable, SmartRecruiters.
Workday, SuccessFactors, Taleo, iCIMS, Recruitee, Teamtailor, Personio and BambooHR are not
supported: they need authentication, vary per tenant, or publish no documented public JSON
endpoint, and guessing at an undocumented one produces a scraper that breaks silently. Those
companies fall back to the career-page scanner.- A JavaScript-rendered careers page returns nothing, and says so as
0 anchors. No headless
browser ships with this pack and none is planned — it would be a Chromium download and a
different product.- Career pages yield titles and URLs, rarely a location, almost never a date. An anchor tag
does not carry one. The reporter pack ages those postings from when it first saw them instead.- robots-disallowed sites return nothing, by design. That is not a bug to work around and
there is no flag that spoofs a browser to get past it.- SmartRecruiters cannot distinguish a wrong slug from an empty board. Their API answers
200with an empty list for a company that does not exist, where the other four return404.
Their identifiers are CamelCase and often suffixed —BoschGroup,McDonaldsCorporation,
Ubisoft2— so copy them out of the board URL. Documented in
reference/finding-your-board-slug.md.- No LinkedIn, Indeed or Glassdoor scraping, and there never will be.
- Scheduled runs use a fixed UTC cron, so your local send time shifts by an hour across
daylight saving. The renderer tells you this in a comment in the file it generates rather than
letting you discover it in March.- This pack fetches; it does not report. No email, no spreadsheet, no state tracking — that is
the companion Job Digest Reporter pack. On its own this one gives you clean JSON and a
scheduler.- Slugs change. A company renames itself or migrates ATS, and the symptom is an
HTTP 404line
for a company that worked last week. Nothing is cached, so fixing the config is the whole fix.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 8 days ago
- Accounts
Google Account
Gmail
- Software
Python
Git
Permissions
Allowed Hosts