How to Integrate bradautomates/claude-video with Other Tools: 4 Deployment Methods

bradautomates/claude-video integrates with any Agent Skills-compatible host—including Claude Code, Codex, Cursor, GitHub Copilot, and Gemini CLI—through a self-contained skill architecture defined in SKILL.md that requires no host-specific code changes.

The bradautomates/claude-video repository implements a host-agnostic video analysis solution built on the Agent Skills framework. By separating the skill contract from runtime scripts, the codebase deploys consistently across 50+ AI assistant platforms without requiring platform-specific path manipulation or dependency management.

Agent Skills Architecture

The integration relies on a declarative structure where skills/watch/SKILL.md serves as the single source of truth. This file declares the skill name (watch) and points the host to the script directory using relative paths. Because the host resolves scripts relative to the SKILL.md location, the same installation works identically across Claude Code, Codex, Cursor, and Gemini CLI.

The architecture consists of these key components:

• skills/watch/SKILL.md — Defines the skill contract, making the /watch command user-invokable and specifying the entry point.
• skills/watch/scripts/watch.py — The main orchestrator that coordinates downloading, frame extraction, and transcription.
• skills/watch/scripts/download.py — Wraps yt-dlp to fetch video metadata and native captions.
• skills/watch/scripts/frames.py — Implements auto-FPS calculation and frame-budget logic using ffmpeg.
• skills/watch/scripts/transcribe.py — Parses VTT captions, deduplicates frames, and orchestrates Whisper transcription.
• skills/watch/scripts/whisper.py — Provides a minimal std-lib client for Groq or OpenAI Whisper APIs.
• skills/watch/scripts/config.py — Reads user-specific configuration from ~/.config/watch/.env.
• skills/watch/scripts/setup.py — Performs pre-flight checks to ensure ffmpeg, yt-dlp, and API keys are present.
• skills/watch/scripts/build-skill.sh — Packages the skill into a .skill bundle for the claude.ai web UI.

Method 1: Claude Code Marketplace (Recommended)

For Claude Code users, install directly from the marketplace to receive automatic updates. The marketplace installer copies the skills/watch/ folder wholesale into Claude Code's skill directory.

/plugin marketplace add bradautomates/claude-video
/plugin install watch@claude-video

Once installed, the /watch slash command becomes available immediately in the Claude Code interface.

Method 2: Universal CLI for Codex, Cursor, Copilot, and Gemini

Any host implementing the Agent Skills CLI supports installation via npx. The -g flag performs a global installation, copying the self-contained skills/watch/ folder into the host's skill path.

npx skills add bradautomates/claude-video -g

This method works identically for Codex, Cursor, GitHub Copilot, Gemini CLI, and over 50 other Agent Skills-compatible hosts. After installation, invoke the skill using:

/watch https://youtu.be/dQw4w9WgXcQ analyze the first 30 seconds

Method 3: Claude.ai Web UI

For the claude.ai web interface, package the skill as a .skill bundle and upload it manually. The repository includes a build script that packages the entire skills/watch/ directory.

bash skills/watch/scripts/build-skill.sh

This produces dist/watch.skill. Upload this file via Settings → Capabilities → Skills in the claude.ai interface. The bundle contains the same SKILL.md contract and scripts, ensuring consistent behavior between the web UI and CLI hosts.

Method 4: Manual Development Setup

For custom automation pipelines or development workflows, clone the repository and symlink the skill folder directly into any host's skill path.

git clone https://github.com/bradautomates/claude-video.git
ln -s "$(pwd)/claude-video/skills/watch" ~/.codex/skills/watch

# Or for Claude Code: ~/.claude/skills/watch

Because the skill relies only on standard command-line tools (yt-dlp, ffmpeg) and pure-stdlib Python, it integrates into any automation pipeline that can invoke Python scripts with those binaries on the PATH.

Programmatic Usage Examples

Beyond slash commands, invoke the entry point directly in custom workflows. The watch.py script functions as a standalone module.

Extract a specific segment with custom detail:

python -m skills.watch.scripts.watch https://youtu.be/abc --start 2:15 --end 2:45 --detail token-burner --out-dir ./analysis

Disable transcription and analyze frames only:

python -m skills.watch.scripts.watch video.mp4 --no-whisper

The script reads configuration from ~/.config/watch/.env using skills/watch/scripts/config.py, allowing API keys and defaults to persist across invocations without hardcoding credentials.

Summary

• bradautomates/claude-video deploys as a self-contained Agent Skill compatible with 50+ hosts including Claude Code, Codex, Cursor, and Gemini CLI.
• The skills/watch/SKILL.md contract ensures host-agnostic behavior by resolving scripts relative to the skill directory.
• Install via Claude Code marketplace (/plugin install), universal CLI (npx skills add -g), web UI bundle (build-skill.sh), or manual symlink.
• The runtime depends only on ffmpeg, yt-dlp, and standard Python, enabling integration into arbitrary automation pipelines.
• Direct invocation via python -m skills.watch.scripts.watch supports custom workflows without the Agent Skills framework.

Frequently Asked Questions

Which AI assistants support bradautomates/claude-video integration?

Any host implementing the Agent Skills framework supports the integration, including Claude Code, OpenAI Codex, Cursor, GitHub Copilot, and Gemini CLI. The SKILL.md contract standardizes the interface across all platforms, allowing the watch skill to function identically regardless of the underlying AI assistant.

What external dependencies are required for manual integration?

The skill requires ffmpeg for frame extraction and yt-dlp for video downloading, both of which must be available on the system PATH. The skills/watch/scripts/setup.py file performs pre-flight validation and can scaffold the ~/.config/watch/.env configuration file. All Python code uses only standard library modules, eliminating pip dependency conflicts.

Can I integrate claude-video without the Agent Skills framework?

Yes. While designed for the Agent Skills ecosystem, skills/watch/scripts/watch.py functions as a standalone Python module. You can invoke it directly via python -m skills.watch.scripts.watch <url> or import its functions into existing Python workflows, provided ffmpeg and yt-dlp are installed on the system.

How do I migrate the skill between different AI hosts?

Migration requires only copying or symlinking the skills/watch/ folder to the new host's skill directory. Because paths in SKILL.md are relative and the Python code uses only std-lib imports, the skill requires no code changes when moving between Claude Code, Cursor, Codex, or other supported platforms. Use the same npx skills add command or manual symlink method on the target host.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}