More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Browser Extension Reverse Engineering

    1

    Find out what an extension really does: map its permission surface from the manifest, trace its background and content-script logic, and follow user data to its destination — verdict with evidence.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Browser Extension Reverse Engineering

    Browser Extension Reverse Engineering

    Example session with this skill installed

    A "dark mode for every site" extension, version 2.1, from the official store. The listing promises a theme toggle and nothing else. Is it doing anything else?

    • Read your context and instructions
    • Compiled the browser extension reverse

    MANIFEST: Permissions: broad host access (plausible for a site-wide theme), storage (settings, fine), tabs (FLAGGED — a theme tool has no tab-management feature), cookies (FLAGGED — no auth feature exists). Two unexplained flags: concerning before code is read; the review focuses on what tabs and cookies are used for.

    ARCHITECTURE: Service worker as coordinator, one content script injected into all pages (the theme injector), a popup with the toggle. No native messaging hosts.

    DATA FLOW: The content script reads the page's text content and sends it to the background worker, which posts it to a third-party analytics domain every 60 seconds. The popup toggle works locally and never touches the network. Storage holds the theme setting only — no tokens.

    RED FLAGS: HIGH — page text exfiltrated to a third-party domain, file and line cited, sensitivity medium (browsing content), destination unrelated to the claimed purpose. MEDIUM — tabs and cookies permissions requested but never used in code. LOW — the analytics domain appears in no privacy note in the listing.

    VERDICT: CONCERNING — the extension delivers its promised theme, but it also ships page content to a third party with no justification. Recommended action: remove, and investigate further only if the vendor explains the flow. The report names the file and line of the exfiltration call, so the vendor's answer can be checked against the code.

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

    About this skill

    An extension's store listing says one thing; its manifest and code say another. This skill runs the analysis in order — listing, manifest, code, data flow — and the verdict comes from the code, not the listing. It starts with the manifest as the declared attack surface: every permission gets a one-line justification or a flag, and two unexplained flags make the extension concerning before a line of code is read.

    Then it maps the architecture (service worker, content scripts, popup, native messaging hosts), follows the data through every message-passing hop, and grades each flow by sensitivity — a single high-sensitivity flow to an unexplained destination is enough to recommend removal. Red flags carry severity and file-and-line evidence: remote code loading, obfuscation in a simple extension, credential-adjacent access, update-time behavior changes. The verdict block (benign, concerning, or malicious) follows written call rules, and a borderline case names what would settle it. A worked example catches a "dark mode" extension exfiltrating page text to a third-party analytics domain.

    • Static reading frames hypotheses about runtime behavior; dormant payloads triggered by date, domain, or remote flag need dynamic observation to confirm.
    • Minified code slows the read but does not block it; deliberate obfuscation is itself a finding.
    • Legitimate broad-permission extensions (password managers, developer tools) need the justification test applied against their claimed purpose, not a universal permission budget.
    • A verdict of concerning on borderline evidence names what would settle it rather than overreaching.

    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

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    Not by itself. Host access is plausible for a site-wide tool and storage holds settings. The skill flags permissions against the claimed purpose — a theme tool asking for debugger access or cookies is the pattern that matters, not the raw count.

    Minification shortens names and strips whitespace but keeps the structure readable; obfuscation hides intent with string arrays, flattened control flow, or encoded payloads. Minification is normal. Obfuscation in a simple extension is a red flag.

    Compare versions. Diff the manifest and the background worker first — malicious extensions often ship clean and turn later, and the update-time change is the highest-value comparison in extension triage.

    New in Security