transcript-video-section-splitter
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple external commands including
ffmpegfor video splitting,ffprobefor metadata extraction, andcondafor running transcription tools. It also references local Python scripts (transcribe.py,split_video_by_sections.py) for processing logic. - [DYNAMIC_EXECUTION]: The skill includes an inline Python script executed via a shell heredoc (`python3
- <<'PY'`) to perform automated validation of the generated video clips.
- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests and processes untrusted data from transcripts and manifest files.
- Ingestion points: Transcript files (
text/source.md) and manifest files (manifest.json) are read into the agent's context. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when processing external content.
- Capability inventory: The skill possesses the ability to execute shell commands (
ffmpeg,ffprobe,conda), run Python interpreters, and perform file system operations. - Sanitization: Content from the manifest (specifically file paths) is passed to
subprocesscalls without explicit sanitization, creating a potential for argument injection if the input is maliciously crafted.
Audit Metadata