name: video-analyzer-openclaw description: Analyze a local video file or video URL with the local video-helper backend from OpenClaw. Use when the user wants to analyze a video, summarize course content, extract highlights and timestamps, generate a mind map, or test browser-compatible playback for uploaded videos.

Video Analyzer For OpenClaw

Use this skill to run the local video-helper backend directly from OpenClaw, without depending on the Codex-side helper skill.

Default flow

From this skill directory, run:

python3 scripts/analyze_video.py "/absolute/path/to/video.mp4" --lang zh --title "视频标题"

For a URL:

python3 scripts/analyze_video.py "https://www.bilibili.com/video/BV1..."

The wrapper will:

1. Create the backend job directly through the API.
2. Wait until the backend reaches blocked or succeeded.
3. For long videos, fetch chunks and auto-build summaries.json.
4. Auto-build plan.json.
5. Submit the plan and poll to completion.

Outputs

Run artifacts are saved under:

data/runs/<projectId>/<jobId>/

Important files:

• chunks.json
• summaries.json
• plan_request.json
• plan.json

Result page

On success, open:

http://localhost:3000/en/projects/<projectId>/results

Configuration

Optional environment overrides:

• OPENCLAW_VIDEO_HELPER_API_URL
• OPENCLAW_VIDEO_HELPER_FRONTEND_URL
• OPENCLAW_VIDEO_HELPER_SOURCE_DIR
• OPENCLAW_VIDEO_HELPER_AUTO_START

The script will auto-load overrides from .env.local or .env in this skill directory.

If the backend is already running on localhost, no extra configuration is needed. If you want this skill to try starting the backend itself, set OPENCLAW_VIDEO_HELPER_SOURCE_DIR to the local video-helper project root and keep OPENCLAW_VIDEO_HELPER_AUTO_START=1.