ClawKit Logo
ClawKitReliability Toolkit
Back to Registry
Official Verified

roughcut

Run RoughCut headlessly on macOS to generate Final Cut Pro (FCPXML) rough-cut timeline variants from a talking-head video — local-first, no media upload.

skill-install — Terminal

Install via CLI (Recommended)

clawhub install openclaw/skills/skills/samergmtm22/roughcut-openclaw
Or

RoughCut (macOS)

This skill lets an OpenClaw agent run RoughCut locally on a user's Mac to produce RoughCut.xml_variants.zip from a raw video file, without uploading media.

Preconditions

  1. Confirm the video is present on local disk and get an absolute file path (example: /Users/alice/Movies/raw.mp4). If the file is on Google Drive/iCloud/Dropbox, ensure it is fully downloaded and accessible as a local file path. Tip: on macOS, you can drag the video file into Terminal to paste its absolute path. If the user is on a different machine than the OpenClaw Mac, agree on a delivery method:

    • Synced folder: a folder that syncs onto the OpenClaw Mac (iCloud Drive, Dropbox, Google Drive Desktop, etc.).
    • Direct download URL: ask the user for a direct HTTPS download link (for example an S3/R2/GCS pre-signed URL), then run RoughCut with --video-url (it downloads into output_root/RoughCut.inputs/ automatically).
  2. Confirm RoughCut repo is present on the same Mac.

  3. If the user enables fluff removal, ensure GEMINI_API_KEY is set in the environment that will run RoughCut. If it is not set, ask the user to provide it and explain it is used only for fluff removal.

Questions To Ask The User

  1. Do you want to remove bad takes? (default: yes)
  2. Do you want to remove fluff/off-topic content? (default: no)
  3. If fluff removal is enabled: what is the video topic/goal? (required)

How To Run

Use repo_root and output_root from OpenClaw config:

bash "$REPO_ROOT/scripts/openclaw/roughcut.sh" \
  --video "$VIDEO_ABS_PATH" \
  --out "$OUTPUT_ROOT" \
  --remove-bad-takes true \
  --remove-fluff false

If the user provides a direct download URL, you can pass it directly (the runner downloads to $OUTPUT_ROOT/RoughCut.inputs/ first):

bash "$REPO_ROOT/scripts/openclaw/roughcut.sh" \
  --video-url "$VIDEO_URL" \
  --out "$OUTPUT_ROOT" \
  --remove-bad-takes true \
  --remove-fluff false

If the URL has no filename (common for pre-signed URLs), include --video-name, and optionally --video-sha256:

bash "$REPO_ROOT/scripts/openclaw/roughcut.sh" \
  --video-url "$VIDEO_URL" \
  --video-name "my_video.mov" \
  --video-sha256 "0123456789abcdef..." \
  --out "$OUTPUT_ROOT" \
  --remove-bad-takes true \
  --remove-fluff false

If fluff removal is enabled:

bash "$REPO_ROOT/scripts/openclaw/roughcut.sh" \
  --video "$VIDEO_ABS_PATH" \
  --out "$OUTPUT_ROOT" \
  --remove-bad-takes true \
  --remove-fluff true \
  --topic "$TOPIC"

The runner prints a single-line JSON result to stdout. On success it includes:

  • xml_variants_zip: absolute path to RoughCut.xml_variants.zip
  • video_path: absolute path to the input video used for processing
  • downloaded_video_path: present only when --video-url was used (where the video was saved)

Metadata

Stars1133
Views0
Updated2026-02-18
View Author Profile
AI Skill Finder

Not sure this is the right skill?

Describe what you want to build — we'll match you to the best skill from 16,000+ options.

Find the right skill
Add to Configuration

Paste this into your clawhub.json to enable this plugin.

{
  "plugins": {
    "official-samergmtm22-roughcut-openclaw": {
      "enabled": true,
      "auto_update": true
    }
  }
}
Safety NoteClawKit audits metadata but not runtime behavior. Use with caution.