dual-coding-designer
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources without explicit boundary markers or sanitization. While the skill's primary function is to generate educational design advice, these ingestion points represent an attack surface where maliciously formatted input could attempt to influence agent behavior.
- Ingestion points:
verbal_content,subject_and_level,visual_constraints,existing_visuals, andstudent_profilesfields inSKILL.md. - Boundary markers: None identified. The prompt interpolates variables directly without delimiters (e.g.,
""") or "ignore instructions" warnings. - Capability inventory: The skill is limited to text generation (design specifications, integration notes, and ASCII sketches) and does not invoke external tools, write files, or perform network operations.
- Sanitization: None identified. The skill relies on the model's internal safety guardrails to handle potentially malicious content within the interpolated fields.
Audit Metadata