bevy-capture
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documents the use of
Mp4FfmpegCliEncoderandMp4FfmpegCliPipeEncoder, which interact with the host system by spawningffmpegas a child process. This behavior is a standard requirement for video encoding and is transparently documented for the user. - [INDIRECT_PROMPT_INJECTION]: The skill provides functionality for writing rendered frames and video files to the local filesystem and spawning external processes. While the examples use hardcoded paths, there is a potential attack surface if developers interpolate untrusted data into file paths or encoder arguments.
- Ingestion points: The skill processes binary image data generated by the Bevy renderer.
- Boundary markers: Not applicable as the skill processes internal application assets rather than text-based prompts.
- Capability inventory: The skill utilizes filesystem write access (
std::fs::File,fs::create_dir_all) and subprocess execution (ffmpeg). - Sanitization: The provided canonical pattern uses static strings for file paths; developers should implement path sanitization if dynamically generating filenames from external inputs.
Audit Metadata