browser-qa
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by ingesting untrusted data (URLs and web page content) which is subsequently processed by automated testing scripts.
- Ingestion points: Browser automation scripts (e.g., in
guides/headless-browser-security-testing-guide.md) navigate to user-specified URLs and interact directly with DOM elements from external websites. - Boundary markers: The scripts do not implement explicit delimiters or system-level instructions to ignore potential commands embedded within the web pages being audited.
- Capability inventory: The skill utilizes
Bash,Agent, and Playwright libraries to perform web navigation, form submission, network interception, and local file evidence capture. - Sanitization: External inputs and fuzzed payloads are interpolated into automation logic without sanitization, which is the intended behavior for security testing but establishes a risk surface for the executing agent.
- [DYNAMIC_EXECUTION]: The skill makes extensive use of dynamic JavaScript execution within the browser's context to monitor application behavior and deliver payloads.
- Scripts in
payloads.mdandguides/network-interception-guide.mdutilizepage.evaluate()andpage.addInitScript()to inject custom logic, such as MutationObservers and prototype pollution checks, into active browser sessions. - [EXTERNAL_DOWNLOADS]: The skill includes standard procedures for installing the Playwright automation framework and its browser binaries from official sources.
- Setup instructions in
guides/headless-browser-security-testing-guide.mdandguides/network-interception-guide.mdfacilitate the download of Node.js and Python packages (playwright,@playwright/test) and browser binaries vianpx playwright install.
Recommendations
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata