Stats
Actions
Tags
From heimdall
Analyze a video file — extract frames, transcribe audio, produce a structured report. Use when the user says "watch this video", "analyze this recording", "screen recording", "bug recording", provides a path ending in .mp4, .mov, .avi, .mkv, .webm, or .m4v, or asks to understand what happens in a video file.
How this skill is triggered — by the user, by Claude, or both
Slash command
/heimdall:watchThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Dispatch the `heimdall` agent to analyze the video file at `$ARGUMENTS`.
Dispatch the heimdall agent to analyze the video file at $ARGUMENTS.
Use the Agent tool to spawn a subagent with this prompt:
Analyze the video file at: $ARGUMENTS
Run the full heimdall workflow:
- Check deps: verify ffmpeg is installed (
which ffmpeg). If missing, tell the user to run/heimdall:setup.- Overview extraction:
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/extract.py "<VIDEO_PATH>" --overview- Read metadata.json, transcript.json, frame_transcript_map.json from the cache directory
- Read frames in batches of 8-10 from the frames/ subdirectory
- If interesting moments found (errors, state transitions), drill down:
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/extract.py "<VIDEO_PATH>" --detail --start <SEC> --end <SEC>- Synthesize and return a structured bug report (or general video analysis)
Do NOT read the video file directly — use extract.py to process it.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub pranavtotla/heimdall --plugin heimdall