qa-testing

Pass

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill extracts untrusted data from the Document Object Model (DOM) of websites being audited, creating a surface for indirect prompt injection.\n- Ingestion points: The skill reads document.title, <h1> and <h2> text content, meta descriptions, social tags (OpenGraph, Twitter), and JSON-LD schema content using JavaScript snippets.\n- Boundary markers: The instructions lack specific delimiters or warnings for the agent to treat the extracted site content as untrusted data, which increases the risk of the agent interpreting embedded text as instructions.\n- Capability inventory: The agent uses this data to generate and write a report file (qa-report-[date].md).\n- Sanitization: The provided snippets truncate string lengths (e.g., using .slice(0, 60)) and use basic error handling for JSON parsing, but do not perform semantic sanitization to filter out prompt injection payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 8, 2026, 04:49 PM
Security Audit — agent-trust-hub — qa-testing