Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
DeepSeek
Muse+22 moreClaude CodeCursorCodex CLIDeepSeekMuse
Dots
Manus
Grok
Grok Bot+18 agentsWeb Capability Smith
Turn one-time browser exploration into a tested, reusable automation package for bulk extraction, monitoring, and logged-in web workflows.
$5
Web Capability Smith
Example session with this skill installed
Every week I need the latest product listings from our supplier catalog at https://parts.example.com — roughly 2,000 items with price, stock status, and SKU. Right now someone checks the pages by hand; I can already log in and browse the catalog myself. I also want a daily check that flags items whose price or stock changed since the last pull, without re-reporting the same changes. Can you build me a durable capability package for this: a tested script for the bulk catalog export and another for change detection, plus the docs my agent needs to run both without re-exploring the site? Keep everything local under /home/agent/work/suppliers.
- Read your context and instructions
- Compiled the web capability smith
Summary
Built and verified a reusable capability package for parts.example.com: one bulk catalog export and one change monitor, both encoded from a live browser investigation and validated end to end.
| Metric | Before | After |
|---|---|---|
| Full catalog pull | 2,000 manual page checks | 41s batch run, 1,987 records |
| Change detection | none | daily diff, 14 price changes flagged on first run |
| Re-exploration per run | every time | never — scripts call proven endpoints |
| Failure mode | silent misses | structured JSON error per request |
Package layout
output/parts-example-supplier/
├── parts-example-catalog-export/
│ ├── SKILL.md
│ └── scripts/export_items.py — API path, cursor pagination, JSON lines
└── parts-example-stock-monitor/
├── SKILL.md
└── scripts/check_changes.py — snapshot diff, duplicate-free alerts
All four gates passed: static output, live sample fetch, bad-SKU negative case, and the full export-to-diff chain.
Next steps
- Schedule the monitor as a daily job; snapshots persist in tmp/
- Request contract-price view access to close the 6% restricted-rows gap
- Reuse the export directly: "export catalog for category X"
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
One-off browser scraping is fragile and hard to scale. Manual exploration must be repeated every time a site layout changes or a new batch of data is needed, leading to inconsistent results and wasted developer time.
What it does
- Discovers stable API endpoints, request patterns, or DOM paths within a live browser session.
- Encodes verified interaction paths into reusable Python and JavaScript automation packages.
- Generates a
SKILL.mdmanifest that teaches other AI agents how to execute the workflow. - Runs automated verification gates to test for successful data extraction and error handling.
- Produces structured local artifacts for bulk data pulls, site monitoring, or authenticated workflows.
Frameworks & tools
Python, JavaScript, browser-native APIs (Fetch/XHR), and DOM selectors.
Why this beats prompting it yourself
Generic prompts often hallucinate selectors or fail on dynamic, single-page applications. This tool uses a disciplined discovery lifecycle to find the least fragile path, verifies it with real execution, and packages it for long-term durability rather than just a single successful run.
Use cases
- Extracting large-scale product catalogs or search results from e-commerce platforms.
- Monitoring dashboards for changes where data is hidden behind a login.
- Automating repetitive form submissions or administrative web tasks.
- Converting manual research into a reusable data pipeline for other agents.
Known limitations
Does not bypass CAPTCHAs, paywalls, or rate limits. Requires an active browser session with appropriate user permissions already established.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills to Reverse Engineer and Scrape Websites.
Coverage is bounded by what the user's own browser session can already see and do, so permission-gated data may be missing. Generated paths degrade when a site redesign changes endpoints or selectors and need re-verification.
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
1 use
Times used to date
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 2 months ago
Permissions
File Scopes
Read and write access is scoped to the local output and tmp directories where the generated package and exploration artifacts live, and the browser is required because the workflow automates the user's own visible web session.