dreamina-canvas-create
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes system-level commands such as uuidgen, tr, and mkdir, and utilizes the dreamina-canvas CLI tool to perform lifecycle management for canvases.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied canvas names and manages state in a local file, which creates a potential surface for indirect injection if inputs are maliciously crafted.
- Ingestion points: User-provided canvas names within the SKILL.md command examples.
- Boundary markers: Not explicitly used to separate user-supplied strings from command parameters.
- Capability inventory: Shell command execution and file system writes to the .state/project-id file.
- Sanitization: The skill enforces that project identifiers are converted to a lowercase UUID format using standard shell utilities.
- Mitigation: Implement strict validation for user-supplied strings and use structured formats like JSON for command arguments where possible to prevent injection.
Audit Metadata