The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/extract_frames.py executes external binaries, specifically ffmpeg, ffprobe, and tesseract, using the subprocess.run() function. These operations are essential for the skill's primary functions, which include video decoding, frame extraction, and OCR analysis. The command execution is implemented using list-based arguments to mitigate the risk of shell command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from video files and transcript JSON files, creating an inherent surface for indirect prompt injection attacks.
Ingestion points: The skill ingests binary video data and JSON-formatted transcripts via the extract_frames.py script.
Boundary markers: The skill's instructions and the generated timeline.json output do not provide specific delimiters or instructions to the agent to treat the OCR-extracted text or speech spans as untrusted content.
Capability inventory: The skill possesses capabilities for file system access (writing frames and data to a work directory) and executing external analysis tools through subprocess calls.
Sanitization: The script performs basic whitespace normalization on OCR output but does not implement security-focused sanitization to detect or neutralize potential prompt injection instructions embedded within the processed text.

video-keyframes