ai-avatar-video
Pass
Audited by Gen Agent Trust Hub on Apr 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documents the use of the infsh CLI tool to interact with AI models for video and audio generation. The execution environment is restricted to this specific tool through the allowed-tools configuration field.\n- [EXTERNAL_DOWNLOADS]: The documentation includes links to installation scripts and related skill packages hosted on the author's official GitHub organization (inference-sh). These are standard resources for the platform's ecosystem.\n- [PROMPT_INJECTION]: The workflow for video dubbing suggests transcribing audio into a text format that is then intended for translation by an LLM. This introduces a surface for indirect prompt injection, as malicious instructions present in the source audio could influence the translation process or subsequent video generation steps.\n
- Ingestion points: External media files processed for transcription via the whisper-large-v3 model.\n
- Boundary markers: None explicitly mentioned in the example workflows to isolate processed text from agent instructions.\n
- Capability inventory: Access to the infsh CLI for running remote AI inference tasks.\n
- Sanitization: The skill does not define specific validation or filtering steps for transcribed text before it is passed to an LLM for translation.
Audit Metadata