The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/clips.py executes ffmpeg and ffprobe using the subprocess module to handle media processing.
Evidence: Subprocess calls are used for linking sources, calculating audio RMS levels, creating low-resolution previews, reframing video, and trimming clips.
Context: These operations use the list-based API which protects against shell injection and are essential to the skill's purpose.
[INDIRECT_PROMPT_INJECTION]: The skill summarizes external video transcripts into shortlist.md for the agent to review.
Ingestion points: Transcripts from arbitrary YouTube videos or local files are parsed and snippets are placed into shortlist.md (processed in scripts/clips.py).
Boundary markers: No explicit markers are used to separate transcript snippets from instructions, although the agent is given specific evaluation criteria.
Capability inventory: The skill has access to file writes, subprocess execution (ffmpeg), and network fetching tools.
Sanitization: The transcript content is extracted without filtering for potential prompt injection patterns.
[EXTERNAL_DOWNLOADS]: The skill uses the yt-dlp tool to download video content from external URLs.
Evidence: Handled via python3 "${CLAUDE_PLUGIN_ROOT}/lib/ai_editor/links.py" fetch "<url>".
Context: This is a standard feature for a video processing skill and uses a well-known, trusted utility.

clips