editorial-portrait-composition-by-s8wxu9
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch and analyze external media and documentation from
goodcase.aiandyoumind.comto extract visual composition traits. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface by ingesting and acting upon data from untrusted external URLs using browser and image analysis tools.
- Ingestion points: External URLs in
references/cases.md(e.g.,goodcase.ai,x.com) are ingested into the agent context during the workflow. - Boundary markers: Instructions include a 'reference contract' (Preserve, Replace, Avoid) and a 'verification' step to maintain structural integrity.
- Capability inventory: The workflow utilizes image, video, and browser tools to inspect remote content and then synthesizes generation prompts.
- Sanitization: There are no explicit mechanisms mentioned to sanitize or filter content retrieved from the external sites before it influences the agent's prompt generation.
Audit Metadata