creating-themes

Pass

Audited by Gen Agent Trust Hub on Jul 26, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is designed to process external site data (HTML and CSS) to replicate themes, which creates a surface for indirect prompt injection.
  • Ingestion points: Captured HTML/CSS from a source site and user-provided prompts (referenced in SKILL.md).
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for the external data.
  • Capability inventory: The agent can write theme files (PHP, JSON, HTML) and execute network downloads for assets.
  • Sanitization: Includes basic rules to sanitize captured line-heights and prevent emoji injection in generated content.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to download font files and logo images from external source CDNs and Google Fonts.
  • Evidence: Instructions in SKILL.md for downloading assets into the theme's assets/ directory.
  • Note: Downloads from Google Fonts are targeting a well-known service, and asset localization from source sites is a privacy and performance best practice.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 26, 2026, 03:02 AM
Security Audit — agent-trust-hub — creating-themes