deck-open-slide-canvas
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill uses the user's message to construct a brief that guides file authoring and tool usage, without providing adequate separation between instructions and data.
- Ingestion points: The user message is incorporated into the brief parameter in
SKILL.md. - Boundary markers: Absent. The prompt does not define clear delimiters to encapsulate the user-controlled brief.
- Capability inventory: Access to file-editing tools and the
open-designMCP server tools for design generation. - Sanitization: Absent. The brief is not filtered or escaped to prevent embedded instructions from being executed by the agent.
Audit Metadata