The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection or safety bypass attempts was found. The skill follows standard instructional patterns for AI agents.\n- [REMOTE_CODE_EXECUTION]: The skill mentions using npx to install additional official skills from the vendor (scenario-labs) if they are missing. This is a recommendation for user interaction rather than an automated execution.\n- [COMMAND_EXECUTION]: The instructions include the use of curl for downloading generated audio assets from Scenario's platform. This represents a legitimate way to handle the output of the generation tools.\n- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface through the processing of user-provided lyrics and descriptions for audio synthesis. However, it does not provide high-privilege tools that could be exploited through this vector, maintaining a low-risk profile.

scenario-audio