More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Fast Browser Use: Command-Line Browser Automation

    1

    Drive a real browser from the command line for scraping and automation: fast DOM extraction, infinite-scroll harvesting, session reuse, snapshots, sitemaps, and job-based collection.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Fast Browser Use: Command-Line Browser Automation

    Fast Browser Use: Command-Line Browser Automation

    Example session with this skill installed

    Watch a startup job board for new backend postings and give me the new ones every morning, with the description and comp range for each.

    • Read your context and instructions
    • Compiled the fast browser use

    Job-board monitor — job built with the skill

    Job definition (written before the first run):

    • Target: the job board's listings URL pattern, found via sitemap first — "the job board" is a site, not a URL.
    • Selector: the job-card selector, verified against a first-page screenshot (real cards, not skeleton loaders) and a one-page extract count that matched what the screenshot showed.
    • Pagination: harvest, 4 scrolls, 1s delay — tuned per site, not guessed.
    • Output: JSON, one record per posting, fields: title, company, posting URL, description, comp range.
    • Dedup key: the posting URL — stable, unique, and the handle for the detail pass.
    • Session: saved once from a headed login, gitignored, scoped to this job; the re-login path is built in because sessions expire, and the failure mode is loud (stop and report), not a silent retry loop.

    Run: nightly via cron. Each run dedups against yesterday's file; only new posting URLs get a detail pass (navigate each, extract description + comp range). Re-scraping every posting every night would waste the site's patience and the job's time. First run found 3 new backend postings; the detail pass returned descriptions and comp ranges for all three, and the screenshot confirmed the selector still matches real cards. Rubric score: 9/10 — one point held back because the comp-range field is missing on some postings; it is recorded as null with the reason, not invented.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Some data only exists behind JavaScript, logins, or infinite scroll — curl cannot reach it. This skill drives a real browser from the command line through Chrome DevTools Protocol: navigate and wait for the content selector (never a fixed sleep), extract structured records into JSON, harvest infinite scroll with tuned scrolls and delays, snapshot the DOM for debugging, screenshot the first page of every job, and discover a site's structure through robots.txt and sitemaps before scraping it. Sessions are saved once from a headed login and replayed headless — treated like credentials, never committed, scoped to the job, with a re-login path built in.

    The core discipline is jobs, not commands: target, selector, pagination, output, and dedup key defined before anything runs, the selector verified against a screenshot and a one-page extract count, and failures loud instead of silent retry loops. A worked example builds a nightly job-board monitor — sitemap first, dedup on the posting URL, detail passes only on new URLs — with a scoring rubric gating scheduled runs at 8/10.

    • It does not defeat serious anti-bot systems — human-emulation flags raise the bar for naive detection only.
    • It does not substitute for an API when one exists; the skill says to prefer the API.
    • Sites can change their markup; the selector verification runs every time for exactly that reason.
    • It respects robots.txt and the site's terms — if a site clearly does not want automation, the skill says to find an API or ask for access, not fight it.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    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 3 days ago

    Permissions

    Terminal / Shell
    Listed3 days ago

    What's inside

    Frequently Asked Questions

    Because fixed sleeps are either too short (flaky) or too long (slow), and the right number changes with the site's load. Waiting for the selector is exact: the job proceeds the moment the content is there.

    Because titles repeat and change; the href is stable, unique, and it is also the handle for fetching details later. A record without a href is a record you cannot revisit.

    Stop and reassess first. The flags might get past this check; they will not get past the next one, and each attempt raises the chance of a ban. The honest options are an API, permission, or a different source.

    New in Workflow Automation