design-describe
Pass
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted visual data (screenshots or videos) that may contain malicious text instructions.
- Ingestion points: External visual data is ingested via the
$ARGUMENTSparameter within the<screenshot>tag inSKILL.md. - Boundary markers: While the input is delimited by XML-style tags, there are no instructions for the model to ignore or sanitize embedded textual commands found within the images or videos.
- Capability inventory: The skill has permissions to create directories and write multiple markdown files (
plan.md,phase-XX-phase-name.md) to the filesystem. - Sanitization: There is no evidence of content validation or filtering for the visual data before it is processed by the multimodal analysis tool.
Audit Metadata