The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing transcribed text from untrusted user media and presenting it to the agent to guide editing decisions.
Ingestion points: Transcribed text and metadata from the Volcengine ASR service are ingested by scripts/transcribe.py and presented to the agent for analysis.
Boundary markers: The script uses output headers such as "【逐段标注】" and "【分段时间戳】" to provide structural separation for the data.
Capability inventory: The skill possesses shell execution capabilities via ffmpeg and file system access in scripts/transcribe.py to perform video modifications.
Sanitization: No specific sanitization or filtering of the transcribed text is performed before it is interpolated into the agent's context.
[COMMAND_EXECUTION]: The skill uses the ffmpeg utility to perform video cutting and concatenation based on transcription data.
Evidence: Use of subprocess.run() in scripts/transcribe.py to execute ffmpeg commands.
Context: This is the primary intended functionality of the skill and is used to process local video files provided by the user.

volc-asr