skills/frankxai/arcanea/visual-proof/Gen Agent Trust Hub

visual-proof

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill facilitates visual verification by automating browser screenshots, providing an evidence-based approach to UI development and reducing the risk of hallucinated UI descriptions.
  • [SAFE]: Input sanitization is implemented for command-line arguments in capture.mjs. Specifically, the label parameter is sanitized to remove special characters and leading dots, preventing directory traversal when creating output folders.
  • [SAFE]: The skill adheres to specific environment constraints, such as using pre-installed browser paths in managed environments and avoiding blocked installation commands.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by rendering external URLs, which could contain instructions targeting the agent's vision capabilities. 1. Ingestion points: The capture.mjs script accepts a --url parameter. 2. Boundary markers: None. 3. Capability inventory: File-write (screenshots to disk via page.screenshot in capture.mjs). 4. Sanitization: None.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 11:06 AM
Security Audit — agent-trust-hub — visual-proof