- Home
- Skills
- Workflow Automation
- Jig: your agent presses the buttons
Works with the AI Agents you already use
Jig: your agent presses the buttons
Your agent writes the endpoint and the form, then describes to you what to click, because it has no browser with your logins in it.
Free
Jig: your agent presses the buttons
Example session with this skill installed
I just added the discount code field to checkout. Open checkout in Jig as the customer role, apply the code SAVE10 to a cart over $50, place the order, and tell me if the total and the server's answer are right. Then make the coupon endpoint return 500 and check what the page shows.
- Read your context and instructions
- Compiled the jig
Checkout with a discount code
Project Shop staging, opened as role customer with role_open.
form_studyread the checkout form: 9 fields, the coupon box is a plain input, Country is a react-select, Delivery date is a date picker.form_fillfilled it in one pass: address, country Poland, delivery 12 Nov, couponSAVE10.- Cart was $64.00. After Apply the page shows $57.60, and
page_networkagrees:
POST /api/coupons/apply 200
{"code":"SAVE10","discount":6.40,"total":57.60}
- Placed the order.
POST /api/orders 201, order number shown on the thank you page.
The happy path works.
Coupon service down
page_mock made /api/coupons/apply answer 500.
- The Apply button spins forever: no error, no timeout.
- The console shows
Uncaught (in promise) TypeError: Cannot read properties of undefined (reading 'total'). - The order can still be placed, at the full price, with no hint the code was ignored.
Filed
finding_file wrote the finding with the steps, the environment, the mocked request and the console error attached.
Suggested fix
Handle a failed coupon request: stop the spinner, say the code could not be applied, keep the button usable. Want me to recheck after the fix?
Connects securely to your tools. The creator never sees your data.
About this skill
Your agent writes the endpoint and the form, then describes to you what to click, because it has no browser with your logins in it. This skill connects it to Jig.
Jig is a desktop workbench for testing web apps. The page under test opens inside its window, so your logins and SSO work as usual, and panels around it read forms, the network, the console and accessibility. With jig-mcp the agent in Claude Code, Cursor, VS Code or Codex works in that window directly.
What the agent can do
- Open the page under test, and open it signed in as one of the project's saved roles: the admin and the customer side by side, each in its own session.
- Read the whole form without a click: every field, its kind, whether it is required, and which widget library drives it. Then fill it in one pass, custom dropdowns and date pickers included.
- Read the network with real response bodies, not just "500", and the console.
- Make the server answer 500, empty, slow or dead for one request, without deploying a branch.
- Take screenshots, with password and card fields covered first.
- File what broke as a finding with the steps, the environment and the server's answer attached.
Safety
Reading is always allowed. Acting passes the same gates as Jig's own chat: the project's consent, its allowed addresses, an isolated tab and an action budget. The folder your editor is open in decides which project it works in.
Price
This skill is free. Jig has a free tier forever, no account needed. Roles, fill sets, network rules, recording and the Claude features are Pro: $99 a year per person, with a full thirty day trial.
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.
Needs the Jig desktop app with Settings, Your editor switched on. The editor must be open in a folder that belongs to a Jig project. Acting on the page passes the project's gates (consent, allowed addresses, action budget). Roles, network rules, fill sets and recording are Jig Pro features; the free tier covers reading, the form map, the network, the audit and findings.
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
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 13 days ago
- Free with an Agensi account
- Software
- Jig desktop app (free tier available)jig-mcp (npm)
Node.js