video-remove-background
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it ingests and processes untrusted external data (video files and URLs) using tools with network and file system capabilities.
- Ingestion points: The
videoparameter in thebria_video_callfunction (defined inreferences/code-examples/bria_video_client.sh) accepts user-provided file paths or remote URLs. - Boundary markers: The skill uses natural language instructions to guide the agent but lacks explicit safety delimiters or "ignore embedded instructions" warnings for the content of the processed videos.
- Capability inventory: The skill uses
curlfor network requests (authentication, file upload, API calls, and polling) and interacts with the local file system to read/write credentials in~/.bria/and to upload local videos. - Sanitization: The shell script performs basic string manipulation and JSON construction but does not include advanced sanitization or filtering of the video content itself before it is sent to the remote API.
Audit Metadata