poster-information-hierarchy-by-iq5ssb
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied inputs and reference data to generate stylized AI prompts, creating a surface for indirect prompt injection.
- Ingestion points: User inputs for subject, audience, and brand assets in SKILL.md; design case data in references/cases.md.
- Boundary markers: No specific delimiters are used to separate user data from internal workflow instructions.
- Capability inventory: Use of browser or image tools to view reference URLs; text generation for AI prompts.
- Sanitization: No explicit sanitization or filtering is applied to the ingested content.
- [DATA_EXFILTRATION]: The skill includes external links to image examples on goodcase.ai and GitHub to inform the creative process. The workflow directs the agent to inspect these links to ensure visual fidelity with the chosen design style. The skill contains explicit safety instructions that prohibit account changes or asset uploads without user approval.
Audit Metadata