byted-las-long-video-understand
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches a manifest and a Python SDK wheel from the vendor's distribution infrastructure at
volces.com. This is a standard procedure for maintaining the tool's dependencies. - [REMOTE_CODE_EXECUTION]: During environment setup, the skill installs a Python package from a remote URL. This execution is scoped to the installation of the vendor's official SDK.
- [INDIRECT_PROMPT_INJECTION]: The skill processes video content provided via URLs or local files, which serves as a potential ingestion point for untrusted data.
- Ingestion points: Video files and URLs supplied to the
video_urlparameter indata.jsonand local file uploads. - Boundary markers: No specific delimiters or "ignore instructions" warnings are utilized in the prompt construction in the documentation.
- Capability inventory: The skill uses the
lasutilCLI to perform video analysis, poll status, and retrieve results. - Sanitization: Input content is passed to the processing engine without local validation or escaping of potential embedded instructions.
Audit Metadata