cowart-image-gen
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from the Cowart canvas environment to influence the behavior of a secondary image generation tool.
- Ingestion points: The skill reads shape properties and metadata using the
get_cowart_selectiontool as described inSKILL.md(Step 1). - Boundary markers: There are no specific delimiters or instruction-isolation warnings provided to the image generation model when interpolating canvas metadata into the prompt (
SKILL.md, Step 4). - Capability inventory: The skill can perform file-write operations to project asset folders and modify canvas state using
insert_cowart_image(SKILL.md, Step 5). - Sanitization: The workflow does not explicitly describe steps to sanitize, filter, or validate the metadata ingested from the canvas before using it in prompt construction.
Audit Metadata