create-brand-theme
Pass
Audited by Gen Agent Trust Hub on Apr 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute JavaScript snippets in a browser environment to scrape computed styles and assets from external websites. It also performs local file system operations using Glob patterns to locate brand assets based on user-provided paths.
- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it ingests untrusted data from external websites (Path A), local folders (Path B), and user-provided URLs (Path C) which is then used to populate themes and influence future generation prompts.
- Ingestion points: Website content, local files, and remote URLs (SKILL.md).
- Boundary markers: Absent; there are no instructions to use delimiters or ignore instructions within the extracted data.
- Capability inventory: The agent has tools to write theme data (update_theme), upload assets (upload_asset), and perform renders (render_html_image, generate_image) in SKILL.md.
- Sanitization: The skill does not describe any validation or sanitization of the extracted CSS, SVG, or text descriptions before storage or use.
- [EXTERNAL_DOWNLOADS]: The skill fetches assets from external URLs and uploads them to vendor-controlled storage at cdn.creativeclaw.co.
Audit Metadata