canvas
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from a whiteboard widget (annotations, instructions, and timecodes) which influences the final generation prompt. This attack surface is well-managed through built-in controls:
- Ingestion points: External data enters via the
get_renoise_whiteboard_revision_intenttool using identifiers provided by the widget. - Boundary markers: The skill requires the agent to present the final prompt to the user in a fenced block for review.
- Capability inventory: The skill has the ability to upload files (
renoise upload) and create paid generation tasks (renoise task create). - Sanitization: A 'hard gate' mechanism ensures that the final prompt string and estimated cost must be explicitly approved by a human user in a separate turn before any paid task is executed.
- [COMMAND_EXECUTION]: The skill interacts with the system via a specific set of Renoise CLI tools. The instructions contain explicit safeguards against command injection and unauthorized file access, such as prohibiting the use of arbitrary local paths and preventing the substitution of sensitive directories (like home or root) for the user-approved project path.
Audit Metadata