Guide · Agent skill
How to use the Interceptor browser and Mac control agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Interceptor is an agent skill that lets an AI agent work in a real Chrome test profile and in native macOS apps, the way a person would. It is useful for checking that a site deploy looks right, reading console and network activity, and reproducing bugs on a page. This guide covers what the skill does, how to install it, how to ask for it in plain language, the requirements, and the safety notes you should read before running it.
Interceptor Browser and Mac Controldanielmiessler/LifeOS · MIT Download the skill Quick start
- 1Download and unzip. Download the Interceptor ZIP from this page and unzip it on your Mac.
- 2Place the skill folder. Move the Interceptor folder into ~/.claude/skills/, or into .claude/skills/ inside a project. Alternatively, upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Install the requirements. Install the interceptor CLI and the Bun runtime, then follow Setup.md to create the interceptor-test Chrome profile.
- 4Grant macOS permissions. Allow Accessibility and Screen Recording for native-app and screenshot work.
- 5Make a natural request. Ask for the outcome, such as verifying a deploy or reproducing a console error, and name the page or app.
- 6Review the evidence. Check the screenshots, console and network state the agent reports before accepting its claim.
What is an agent skill, and what does Interceptor add?
An agent skill is a folder of instructions and scripts that an AI agent loads when a request matches the skill's purpose. Interceptor is one of these, and its job is browser and desktop interaction.
Once installed, the agent can open a separate, isolated Chrome test profile and interact with web pages: reading page structure, clicking, filling forms, checking console and network activity, and taking screenshots to confirm a deploy or reproduce a bug. It can also read and drive native macOS apps through their accessibility interface, and record and replay workflows.
- Verify that a deployed site looks and behaves correctly at desktop and mobile widths
- Reproduce visual, animation or console-error bugs in a real browser session
- Inspect a page behind a login inside a sandboxed browser profile
- Read and drive native macOS apps through accessibility handles
Install the Interceptor skill
Download the Interceptor ZIP from this page and unzip it. Place the Interceptor folder in your personal skills directory at ~/.claude/skills/, or in a project's .claude/skills/ folder if you want it limited to one project.
Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: the .claude/skills/ folder inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
Meet the requirements before the first run
Interceptor needs macOS with the interceptor CLI installed, Google Chrome with a dedicated interceptor-test profile, and the Bun runtime, which runs the ContextLoad loader and TypeScript tools. For native-app and screenshot work, macOS also needs Accessibility and Screen Recording permissions.
First-time setup is described in the skill's Setup.md file, which covers the program install, the test Chrome profile and extension, and preferences.env.
- macOS with the interceptor CLI installed
- Google Chrome with a dedicated interceptor-test profile
- Bun runtime
- Accessibility and Screen Recording permissions for native-app and screenshot work
Trigger the skill with natural requests
You do not need special syntax. The skill's trigger phrases include verify deploy, confirm UI, screenshot verification, debug web, visual check, animation bug, console errors, JS errors, network traffic, HAR export, blank-page debug, what is happening on the page, authenticated page, reproduce bug, and drive native app. Describe the outcome you want and name the page or app.
Requests that the skill does not cover well are pixel-only GUI work, which belongs to a separate Computer Use tool, and crawling or scraping, which belongs to other tools.
- "Verify the deploy of the pricing page at desktop and mobile widths and show me the screenshots."
- "Check the console and network errors on the checkout page and tell me what fails."
- "Reproduce the menu flicker bug on the staging site and capture the state."
- "Open my interceptor-test profile and log in to the staging dashboard, then read the account settings."
Read the evidence the skill returns
The skill is built to require evidence from the real page before the agent claims something works. A verified claim should cite a structured read of the page, console and network state, and a screenshot the agent actually read. When a screenshot is requested, the skill expects one per width the request names, usually 1440 and 390, and checks that each is the requested page.
If Interceptor could not deliver evidence, the report should say verification was not captured rather than substitute other evidence. Treat any report that says a page is fine without screenshots or page state as incomplete.
- Structured page read
- Console and network state
- Screenshot per requested width
- Native-app frontmost app checked before and after
Safety: scripts, network and your browser profile
Interceptor runs scripts and uses the network. The skill is designed to run in an isolated test context pinned in preferences.env, and to leave your Default Chrome profile, windows and focus untouched. Its preflight gate must pass before any browser action, and a gate failure ends the run instead of falling back to another profile.
Do not point it at your personal Chrome profile or at accounts you have not set up for testing. The isolation is meant to protect real sessions, but a misconfigured run could still touch them. Secrets should stay outside the skill, and the skill hands OAuth or consent screens to you to approve.
- Use only the dedicated interceptor-test profile
- Keep secrets out of the skill folder
- Expect to approve OAuth or consent screens yourself
- Native-app work should leave your frontmost app unchanged
What people use it for
- Check a website deploy at 1440 and 390 pixel widths and confirm the layout and requested elements.
- Capture console errors and failed network requests while reproducing a checkout bug.
- Inspect a page behind a login in the sandboxed test profile.
- Read the accessibility tree of a native macOS app to find a button and report its state.
Frequently asked questions
How do I install the Interceptor agent skill?
Download the Interceptor ZIP from this page and unzip it. Place the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder, or upload it in the Claude apps under Settings → Capabilities → Skills.
What do I need to run Interceptor?
You need macOS with the interceptor CLI installed, Google Chrome with a dedicated interceptor-test profile, and the Bun runtime. Native-app and screenshot work also needs Accessibility and Screen Recording permissions.
Is the Interceptor skill free?
The skill is released under the MIT licence, as listed on this page. Check the licence file in the repository for the full terms.
Can Interceptor be used for web scraping?
No. The skill says it is not for simple scraping or large-scale crawling, and it is slower and riskier than purpose-built scrapers. Use other tools for those jobs.
Will Interceptor touch my personal Chrome profile?
It is designed to work only in an isolated interceptor-test profile and to leave your Default profile untouched. Do not use it on your personal profile or accounts you have not set up for testing, because a misconfigured run could still reach real sessions.
How is Interceptor different from ComputerUse?
Interceptor works through the browser's page structure and macOS accessibility handles, and it reads console and network state. Pixel-only GUI automation is directed to a separate Computer Use tool.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Interceptor suits developers, QA and front-end engineers who need an agent to check deploys, reproduce browser bugs or inspect a page behind a login in a sandboxed profile, and who are comfortable setting up a dedicated test Chrome profile. If you need scraping, crawling or pixel-only desktop automation, pick a purpose-built tool instead.
Download the skill