talking-head-production
Pass
Audited by Gen Agent Trust Hub on Apr 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references installation documentation for its required CLI tool, which is hosted on the vendor's GitHub repository.
- [COMMAND_EXECUTION]: The skill uses the
beltCLI tool to perform media generation and account authentication, which is permitted by the skill's manifest configuration. - [REMOTE_CODE_EXECUTION]: Documentation provides instructions for installing supplementary skill modules using the
npx skills addcommand, which downloads content from the vendor's official repository. - [PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection via user-provided text inputs for audio and video generation.
- Ingestion points: Text prompts provided in JSON payloads to the
belt app runcommand for generating narration or portraits (SKILL.md). - Boundary markers: Input data is encapsulated within structured JSON objects to separate instructions from data.
- Capability inventory: The skill utilizes the
beltCLI tool for media synthesis and processing. - Sanitization: Standard tool-level validation is expected; no custom sanitization steps are defined in the instructions.
Audit Metadata