design-concepts
Pass
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Retrieves design inspiration and visual references from well-known community galleries including Dribbble, Behance, and Awwwards. These are treated as safe sources for design research.
- [EXTERNAL_DOWNLOADS]: Identifies typography recommendations and fetches font resources from Google Fonts to style UI mockups, which is standard practice for digital design.
- [PROMPT_INJECTION]: Uses emphasizing language such as 'CRITICAL' and 'IMPORTANT' to guide the agent through specific methodology steps and ensure consistency in deliverables. These directives are intended for process control and quality assurance.
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing data from external websites.
- Ingestion points: Data enters the context via
web_searchresults andweb_fetchcontent from third-party design platforms like Dribbble and Behance (SKILL.md). - Boundary markers: The instructions do not define specific delimiters or 'ignore' instructions for the ingested content to prevent accidental obedience to embedded text.
- Capability inventory: Includes
viewfor file reading,web_searchandweb_fetchfor network access, and implied file writing for project organization (SKILL.md). - Sanitization: There is no explicit requirement for the validation or sanitization of external content before its use in generating UI code.
- [COMMAND_EXECUTION]: Recommends organizing generated design assets within structured directory hierarchies on the local file system, which involves standard file management operations.
Audit Metadata