web-testing

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes a PowerShell script (scripts/test-scaffold.ps1) used to automate the creation of test files and directory structures. While the script operates on local templates and parameters, the use of executable scripts for file generation is a capability to be used with caution.\n- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it is designed to browse, read, and interact with external web content which is inherently untrusted.\n
  • Ingestion points: The skill retrieves data from arbitrary URLs using Playwright's page.goto() and DevTools' navigatePage() as seen in SKILL.md.\n
  • Boundary markers: The provided implementation examples do not include boundary markers or system instructions to disregard potential commands embedded within the processed web data.\n
  • Capability inventory: The skill provides access to browser-level execution environments via page.evaluate() and chrome.evaluateScript(), as well as interaction tools like click and fill, which could be manipulated by malicious content on a target website.\n
  • Sanitization: No mechanisms for sanitizing or validating external web content are demonstrated in the skill's logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 11:23 PM