The Agent Skills Directory

[COMMAND_EXECUTION]: The orchestrator scripts process_video.py and remap_chapters.py utilize the subprocess module to run FFmpeg and execute helper Python scripts like transcribe_video.py and suggest_chapters.py. All command invocations use argument lists rather than raw shell strings, which is a secure implementation that prevents shell injection attacks.
[INDIRECT_PROMPT_INJECTION]: The skill processes video transcripts (external data) to generate chapter titles, descriptions, and Markdown documentation. This introduces a surface for indirect prompt injection where adversarial audio content could be transcribed and later influence the agent when it reads the generated files.
Ingestion points: The scripts read from transcript.json, which is generated from external video content via the Whisper API.
Boundary markers: There are no specific delimiters or warnings used in the generated documentation to segregate external transcript data from agent instructions.
Capability inventory: The skill possesses file-writing capabilities and uses subprocess.run to interact with system tools and other scripts.
Sanitization: The transcript content is processed for remapping and documentation without specific filtering or sanitization for prompt injection patterns.

video-full-process