soundcraft
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions rely on the execution of the
soundcraft-cliandsoundcraftbinaries on the local host to interact with the audio engine and perform UI automation. - [EXTERNAL_DOWNLOADS]: The protocol documentation identifies
github.com/storytold/soundcraftas the upstream source for its control protocol and command definitions. - [INDIRECT_PROMPT_INJECTION]: The skill provides the agent with capabilities to read and process metadata from external audio sessions, creating a surface for indirect instructions.
- Ingestion points: The agent is instructed to use
session.inspectandui.inspectto read track names, clip IDs, markers, and audio routing information from the active SoundCraft session. - Boundary markers: The instructions lack explicit delimiters or warnings to the agent to treat strings found in session metadata (like track names or marker labels) as untrusted data.
- Capability inventory: The skill allows for local command execution via
soundcraft-cli, arbitrary file writes through theui.screenshotpath parameter, and interactive tool control via the TCP loopback bridge. - Sanitization: No sanitization or filtering logic is prescribed for the data retrieved from
session.inspectbefore it is incorporated into the agent's reasoning context.
Audit Metadata