6missedcalls/video-editing-skill

---

Why This Skill?

AI agents are great at writing code — but when you ask them to edit a video, they reach for bloated Python libraries, spin up runtimes, and still can’t chain operations together.

• Pure Bash + FFmpeg — no Python runtimes, no package managers, no build steps
• Composable pipeline — chain trim, jump cut, speed, caption, and overlay in a single command
• 3 caption styles — Hormozi, standard, and minimal — burned directly into the video
• Smart silence removal — auto-detects dead air with configurable thresholds
• Whisper transcription — generates SRT files locally, no API calls, no cloud dependency
• Multi-agent compatible — built for OpenClaw, Claude Code, and Codex; used and tested with OpenClaw
• Works standalone — every script runs independently, with or without an AI agent

---

Table of Contents

• Requirements
• Quick Start
• Capabilities
• Full Pipeline
• Scripts Reference
• Configuration
• Troubleshooting
• Project Structure
• Agent Compatibility
• Contributing
• License

---

Requirements

You must have the following installed on your machine before using this skill.

Install FFmpeg

# macOS (Homebrew)
brew install ffmpeg

# Ubuntu / Debian
sudo apt install ffmpeg

# Arch Linux
sudo pacman -S ffmpeg

# Windows (Chocolatey)
choco install ffmpeg

Install Whisper (optional — only needed for captions)

pip install openai-whisper

Whisper runs locally on your machine. No API keys, no cloud calls, no data leaves your computer.

Verify Installation

ffmpeg -version     # should print version info
ffprobe -version    # should print version info
whisper --help      # should print usage (only if you need captions)

Or run the onboard script to check everything at once:

./scripts/onboard.sh --check

---

Quick Start

Install the Skill

The onboard script checks dependencies, makes scripts executable, and registers the skill with OpenClaw and Claude Code:

./scripts/onboard.sh

Run Your First Edit

# Remove silence from a video
bash scripts/jumpcut.sh ~/video.mp4

# Trim to a specific range
bash scripts/trim.sh ~/video.mp4 --start 00:01:00 --end 00:05:00

# Full pipeline: trim, remove silence, add captions, speed up
bash scripts/edit.sh ~/video.mp4 \
  --trim-start 00:00:10 --trim-end 00:10:00 \
  --jumpcut \
  --caption --caption-style hormozi \
  --speed 1.25 \
  --output final.mp4

That’s it. No install step, no config files, no build process.

---

Capabilities

Trimming

Cut video to specific start/end timestamps or duration. Uses fast codec copying — no re-encoding.

scripts/trim.sh video.mp4 --start 00:01:30 --end 00:05:00

Jump Cuts (Silence Removal)

Auto-detects and removes silent sections. Reports time saved and percentage removed.

scripts/jumpcut.sh video.mp4 --threshold -30 --duration 0.5 --padding 0.1

Captions

Two-step process: transcribe audio with Whisper, then burn subtitles into the video.

Step 1 — Transcribe:

scripts/transcribe.sh video.mp4 --model base --language en

Step 2 — Burn captions:

scripts/caption.sh video.mp4 video.srt --style hormozi

Caption Styles

Text Overlays

Add positioned text at specific timestamps with configurable appearance.

scripts/overlay-text.sh video.mp4 \
  --text "Subscribe!" \
  --start 00:01:00 --end 00:01:05 \
  --position bottom-right \
  --fontsize 48 --color white

Positions: center • top • bottom • top-left • top-right • bottom-left • bottom-right

Speed Changes

Adjust playback speed with pitch-corrected audio. Handles extreme values via automatic filter chaining.

scripts/edit.sh video.mp4 --speed 1.5

---

Full Pipeline

The edit.sh orchestrator chains multiple operations in a single command. Operations execute in this order:

trim → jump cut → speed → caption → overlay

scripts/edit.sh video.mp4 \
  --trim-start 00:00:10 --trim-end 00:10:00 \
  --jumpcut \
  --jumpcut-threshold -25 \
  --caption --caption-style hormozi \
  --speed 1.25 \
  --overlay-text "Like & Subscribe" \
  --overlay-start 00:01:00 --overlay-end 00:01:05 \
  --output final.mp4

Intermediate files are automatically cleaned up. The pipeline reports output file path, duration, and file size on completion.

---

Scripts Reference

All scripts accept --help and include usage documentation in their headers.

---

Configuration

No config files, no JSON manifests. Every option is a command-line flag with sensible defaults.

---

Troubleshooting

---

Project Structure

video-editing-skill/
  SKILL.md                  # Skill metadata for AI agents
  README.md                 # This file
  LICENSE                   # MIT License
  scripts/
    onboard.sh              # Setup: dependency check + skill install
    edit.sh                 # Pipeline orchestrator
    trim.sh                 # Video trimming
    jumpcut.sh              # Silence removal
    transcribe.sh           # Whisper transcription
    caption.sh              # Caption burning (3 styles)
    overlay-text.sh         # Text overlay insertion

---

Agent Compatibility

Built for OpenClaw, Claude Code, and Codex. Used and tested with OpenClaw.

This skill follows the Agent Skills Protocol via SKILL.md and works with any AI agent that can execute shell commands, including:

OpenClaw • Claude Code • Codex • Cursor • Windsurf • Cline • Roo Code • Goose • Continue • Kilo • Amp • and more

Natural language examples:

Edit my video at ~/video.mp4 — remove silence, add Hormozi-style captions, and speed it up 1.25x

Trim ~/interview.mp4 from 00:02:00 to 00:15:00 and add standard captions

Add a "Subscribe!" text overlay at the 1 minute mark in ~/video.mp4

The agent reads SKILL.md, identifies the operations needed, and calls the appropriate scripts.

---

Contributing

Contributions welcome! Please:

1. Fork the repo
2. Create a feature branch (git checkout -b feat/my-feature)
3. Test your changes against real video files
4. Submit a PR

---

License

MIT