dogfood

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it incorporates untrusted data from external websites into the agent's decision-making process.
  • Ingestion points: Untrusted content is retrieved from the target URL via browser_snapshot (DOM content), browser_vision (visual analysis of screenshots), and browser_console (JavaScript console logs) as described in Phase 2 of SKILL.md.
  • Boundary markers: The workflow does not include boundary markers or instructions for the agent to ignore potentially malicious commands embedded in the website's data.
  • Capability inventory: The skill grants the agent access to interaction tools with significant side effects, including browser_click, browser_type, browser_press, and browser_navigate as defined in the Prerequisites and Tools Reference sections.
  • Sanitization: There is no mechanism to sanitize or validate the content retrieved from the browser before it is processed by the LLM.
  • [NO_CODE]: This skill contains no executable source code or binaries, consisting entirely of Markdown instructions, report templates, and taxonomy references.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 05:55 PM