proofread
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation describes the installation of vendor-specific packages, including
soniloandsonilo-cli, through standard package managers (npm and pip). These resources are managed by the skill's author, sonilo-ai. - [INDIRECT_PROMPT_INJECTION]: The skill processes external video and audio data which may contain embedded instructions designed to influence the agent's behavior during transcription or translation.
- Ingestion points: External media provided through
video_pathorvideo_urlparameters inSKILL.md. - Boundary markers: There are no explicit delimiters or boundary markers mentioned to isolate transcribed content from potential instructions.
- Capability inventory: The skill utilizes the
Writetool to save subtitle files to the local filesystem and theBashtool for CLI interactions. - Sanitization: No technical sanitization of the transcribed text is described; however, the instructions implement a manual mitigation by requiring the agent to present files for user review and approval before proceeding to any subsequent steps.
Audit Metadata