design-audit

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted data from external sources (URLs, screenshots, and codebase) to perform UI audits, creating a surface for indirect prompt injection where malicious instructions on a scanned page could influence agent behavior.
  • Ingestion points: Uses browser automation for live URLs, vision tools for processing images, and file tools for reading codebase components.
  • Boundary markers: There are no explicit instructions or delimiters defined to separate untrusted external content from the agent's system instructions.
  • Capability inventory: Browser automation for network access, file reading for context, and file writing to the local ./design/ directory.
  • Sanitization: The workflow does not specify any sanitization, filtering, or validation of content fetched from external targets before it is processed by the AI.
  • [DATA_EXFILTRATION]: The skill performs outbound network operations via browser automation and web-fetch tools to access arbitrary URLs provided by the user or found in the codebase.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 04:37 AM
Security Audit — agent-trust-hub — design-audit