product-studio-visual
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references and directs the agent to interact with multiple external domains for visual evidence and prompt references, including
goodcase.ai,liblib.art,douyin.com, andx.com. The workflow instructions specify: 'Inspect the anchor's finished media with an available image, video, or browser tool,' which involves the agent fetching data from these third-party sources. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external content (prompt excerpts and case summaries) from
references/cases.md. This creates a surface for indirect prompt injection where instructions within the example prompts could influence agent behavior. - Ingestion points: The agent reads the local
references/cases.mdfile containing excerpts from various creators. - Boundary markers: The skill implements a 'Required evidence gate' and a strict 'Output contract' (Preserve/Replace/Avoid) to force the agent to categorize and filter the information.
- Capability inventory: The agent is encouraged to use image, video, or browser tools to inspect the external URLs linked in the evidence.
- Sanitization: There is no explicit sanitization of the ingested prompt excerpts, but the 'Safety and attribution' section provides explicit negative constraints against unauthorized publishing or account changes.
Audit Metadata