The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.run to call ffmpeg for extracting audio segments and performing speech activity detection (VAD) in tips/口误识别方法论.md and tips/转录最佳实践.md.
[EXTERNAL_DOWNLOADS]: The skill requires installing external Python packages funasr and modelscope and downloads pre-trained AI models from ModelScope.
[DYNAMIC_EXECUTION]: The skill executes Python code for character-level tokenization and VAD detection at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted video data for transcription and review.
Ingestion points: Processes user-provided *.mp4 video files to generate _transcript.json and _审查稿.md (SKILL.md).
Boundary markers: Absent; instructions lack delimiters or warnings to ignore malicious instructions within transcripts.
Capability inventory: Shell command execution via subprocess.run and file system operations like os.remove.
Sanitization: No sanitization of the transcribed speech text is performed before it is used for automated tasks.

videocut-clip-oral