ideate
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to ingest a wide variety of external and untrusted inputs, including product URLs, Figma files, screenshots, and codebase paths, to inform its design generations. This ingestion process could potentially expose the agent to adversarial instructions embedded within those external assets.
- Ingestion points: Inputs enter the context through URLs, Figma references, screenshots, and local codebase paths specified in
SKILL.md. - Boundary markers: The instructions do not define explicit delimiters or instructions to ignore embedded commands within the ingested visual or text data.
- Capability inventory: The skill utilizes the
Image Gentool,get-contexttool, and has read access to the local file system (e.g.,storybook/,app/folders). - Sanitization: No explicit sanitization or validation of the content within the reference files is mentioned before processing.
- Dynamic Execution Pattern: The skill instructions specify loading a
user-contextfile and running a 'preflight script' when shell access is available. This pattern involves the execution of local scripts to set up the environment, which is a common practice for environment initialization but warrants review to ensure the scripts performed are limited to expected setup tasks.
Audit Metadata