How to use the Browser Extension Reverse Engineering agent skill

What does the skill make the agent do?

An agent skill is a folder of instructions and reference notes that an AI agent loads when a task matches it. This one gives the agent a step-by-step method for analysing a browser extension. It first confirms the target is a browser extension (an unpacked folder or a .crx or .xpi file) rather than ordinary web page JavaScript.

After unpacking, the agent reads the manifest to list what the extension is allowed to access. It then looks at background scripts, content scripts and network or storage hooks to work out how the extension handles data, signatures or encryption. The skill also covers spotting over-broad permissions and investigating malicious or supply-chain-poisoned extensions.

• Unpack the extension and read manifest.json (permissions, host_permissions, background, content_scripts)
• Flag over-broad permissions such as <all_urls>, webRequest and debugger
• Trace the background service worker or background scripts and content script injection points
• Review chrome.storage and IndexedDB use, and message passing via runtime.sendMessage
• Load the unpacked folder in developer mode and inspect errors on chrome://extensions
• Refer to the bundled analysis notes covering host_permissions, webRequestBlocking, nativeMessaging and externally_connectable

How do you install the skill?

Download the skill ZIP from the skill's page in this directory, then unzip it. You can install it for yourself or for a single project.

For a personal install, unzip the skill into ~/.claude/skills/. For a project, unzip it into the project's .claude/skills/ folder. Alternatively, if you use the Claude apps, you can upload the skill under Settings → Capabilities → Skills.

• Personal: place the skill folder in ~/.claude/skills/
• Project only: place the skill folder in the project's .claude/skills/
• Claude apps: upload the skill under Settings → Capabilities → Skills

How do you trigger the skill with a natural request?

You do not need a special command. Describe the task in plain language and mention the browser extension, so the agent matches it to this skill. Give it the extension folder or the .crx or .xpi file, and say what you want to learn.

Good requests name the goal, such as checking what data an extension can read, or working out how it signs its requests. If the target is a normal web page script rather than an extension, the skill itself points you to the separate js-reverse skill.

• "Check the permissions in this unpacked Chrome extension and tell me which ones are over-broad."
• "Find the background service worker in this .crx and explain what it does with stored credentials."
• "Audit this Firefox .xpi for network request modification."

What does a typical analysis look like?

A common run starts with the manifest, because it lists the permissions and entry points that everything else depends on. The agent then follows those entry points into the background worker and content scripts, and checks network and storage hooks along the way.

The skill's reference notes highlight risk signals to look for. Host permissions of <all_urls> can read and write any site. webRequestBlocking can rewrite traffic in a man-in-the-middle style. nativeMessaging connects the browser to a program on the local machine. externally_connectable lets web pages drive the extension. For Manifest V3 extensions, the notes point to service worker lifecycle and declarativeNetRequest.

What do you need before you start?

The skill expects Chrome DevTools or a Chromium-based browser for debugging, plus jq and an archive tool to unpack .crx and .xpi files. These are needed for the core workflow.

• Required: Chrome DevTools or a Chromium-based browser
• Required: jq and an archive tool for unpacking .crx and .xpi files
• Optional: the js-reverse toolchain for deeper JavaScript analysis
• Optional: Frida or Chrome DevTools Protocol tooling (jshookmcp) for dynamic analysis
• Optional: YARA for writing malicious-extension rules

What are the safety notes?

The skill is framed around authorised testing. Only inspect extensions you have permission to analyse. The skill does not run scripts and does not use the network on its own, according to its listing. Its optional tools may do more, so check what each one does before you run it.

The output is a technical workflow and set of findings, not a verdict on whether an extension is safe. Recovering credential or signature logic can be misused, so treat any recovered secrets with care and do not use the method against systems you are not authorised to test.

• Work only on extensions you are authorised to inspect
• The skill runs no scripts and uses no network by itself
• Optional tools (Frida, CDP tooling, YARA) are separate; review them before use
• Handle any recovered keys or tokens as sensitive material

What tips help get better results?

Start with the manifest and permissions, since they narrow the search. Keep the unpacked folder separate from the original file so you can compare. Use developer mode and chrome://extensions to catch load errors before reading deeper code.

The skill's instructions reference a field-journal precedent file and related skills for JavaScript and malware analysis. Those files are not part of the listed package, so do not count on them being present.

• Read manifest.json first and list permissions before anything else
• Load the unpacked folder in developer mode and check chrome://extensions for errors
• Attach DevTools to the service worker to watch its behaviour
• Ask for one question at a time, such as permissions, then data flow