The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands to perform video and audio processing.
Invokes ffmpeg for video trimming, audio extraction, and rendering.
Uses whisper for speech-to-text transcription.
Executes local Python scripts (analyze.py, build_pan.py, etc.) for specialized processing logic.
[INDIRECT_PROMPT_INJECTION]: The skill interprets untrusted content derived from audio transcripts to identify highlight moments.
Ingestion points: Transcripts generated by whisper are read by the agent in SKILL.md (Step 1) to identify "funny" segments.
Boundary markers: Absent; the agent processes the transcript directly.
Capability inventory: Shell command execution via ffmpeg and whisper, and file system writes in /tmp/clipify/.
Sanitization: Absent; however, the workflow includes a manual checkpoint where the agent proposes candidates to the user for confirmation before proceeding with rendering.

clipify