ui-receipts

Pass

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill interacts with external, untrusted websites to perform UI audits. This creates a potential surface where malicious content on a target page could attempt to influence the agent's report or audit logic.
  • Ingestion points: The scripts/probe.py script extracts text content, console messages, and accessibility attributes from remote URLs provided by the user.
  • Boundary markers: The skill provides a specific report template (# UI receipts: ...) and requires evidence ('receipts') for every finding, which helps ensure the agent remains grounded in observed data rather than potentially malicious text.
  • Capability inventory: The agent uses results to generate a critique. It does not perform sensitive system modifications or network exfiltration based on the page content.
  • Sanitization: The skill relies on the agent's internal safety filters to ignore deceptive prompts found on target websites.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of standard development tools for browser automation.
  • The documentation instructs the user to install playwright via pip and download the Chromium browser. These resources are provided by Microsoft, a trusted vendor, and are necessary for the skill's primary function of visual auditing.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 31, 2026, 05:16 PM