Gobbler Skills¶

Skills are reusable, filesystem-based instructions that teach Claude how to use the gobbler CLI to convert content to markdown.

What are Skills?¶

Skills are markdown files (SKILL.md) with YAML frontmatter that Claude discovers and loads on-demand. Each skill contains:

• Metadata (frontmatter) - name and description that tell Claude when to use the skill
• Instructions - CLI commands, examples, and options for completing tasks

Skills use progressive disclosure: Claude only loads ~100 tokens of metadata at startup. The full instructions are read only when the skill is triggered.

Skill Structure¶

Each Gobbler skill is a SKILL.md file:

skills/gobbler-youtube/
└── SKILL.md           # Instructions with YAML frontmatter

Example SKILL.md:

---
name: gobbler-youtube
description: Transcribe YouTube videos to markdown. Use when user wants to get transcripts from YouTube.
---

# Gobbler YouTube

Convert YouTube videos to markdown transcripts.

## Transcribe Video

```bash
# Basic transcription
gobbler youtube "https://youtube.com/watch?v=VIDEO_ID"

# With timestamps
gobbler youtube "https://youtube.com/watch?v=VIDEO_ID" --timestamps

# Save to file
gobbler youtube "https://youtube.com/watch?v=VIDEO_ID" -o transcript.md

## How Claude Uses Skills

1. **Discovery** - Claude sees skill metadata in its system prompt
2. **Trigger** - When your request matches a skill's description, Claude reads `SKILL.md`
3. **Execute** - Claude runs the `gobbler` CLI commands from the instructions
4. **Output** - CLI returns markdown that Claude can use or save

Skills work with Claude Code, Claude Desktop, and OpenCode. They're discovered from:
- `skills/gobbler-*/SKILL.md` (in the Gobbler repo)
- `.claude/skills/*/SKILL.md` (Claude Code compatible)

## Available Skills

| Skill | Description | CLI Command |
|-------|-------------|-------------|
| `gobbler-youtube` | YouTube transcription | `gobbler youtube URL` |
| `gobbler-audio` | Audio/video transcription | `gobbler audio FILE` |
| `gobbler-document` | PDF, DOCX, PPTX, XLSX conversion | `gobbler document FILE` |
| `gobbler-webpage` | Web page to markdown | `gobbler webpage URL` |
| `gobbler-browser` | Browser control via extension | `gobbler browser ...` |
| `gobbler-notebooklm` | NotebookLM interaction | `gobbler notebooklm ...` |
| `gobbler-chatgpt` | ChatGPT via browser | `gobbler chatgpt ...` |
| `gobbler-claude` | Claude.ai via browser | `gobbler claude ...` |
| `gobbler-gemini` | Gemini via browser | `gobbler gemini ...` |
| `gobbler-setup` | Installation and troubleshooting | Various |

## Skill Reference

### gobbler-youtube

Transcribe YouTube videos to markdown.

```bash
gobbler youtube "https://youtube.com/watch?v=VIDEO_ID"
gobbler youtube "URL" --timestamps           # Include timestamps
gobbler youtube "URL" --language es          # Specific language
gobbler youtube "URL" -o transcript.md       # Save to file

gobbler-audio¶

Transcribe audio and video files using Whisper.

gobbler audio /path/to/audio.mp3
gobbler audio /path/to/video.mp4 --model medium    # Larger model
gobbler audio /path/to/audio.mp3 --timestamps      # With timestamps
gobbler audio /path/to/audio.mp3 -o transcript.md  # Save to file

Models: tiny, base, small (default), medium, large

gobbler-document¶

Convert documents to markdown using Docling.

Requires: Docling Docker container (make start-docker)

gobbler document /path/to/file.pdf
gobbler document /path/to/file.docx --no-ocr    # Skip OCR (faster)
gobbler document /path/to/file.pdf -o output.md # Save to file

Supported formats: PDF, DOCX, PPTX, XLSX

gobbler-webpage¶

Convert web pages to markdown using Crawl4AI.

Requires: Crawl4AI Docker container (make start-docker)

gobbler webpage "https://example.com"
gobbler webpage "https://example.com" -o page.md

gobbler-browser¶

Control browser tabs via the Gobbler extension.

Requires: Browser extension installed, tabs in "Gobbler" group

gobbler browser status              # Check connection
gobbler browser list                # List controlled tabs
gobbler browser extract             # Extract current page
gobbler browser extract -o page.md  # Save to file

gobbler-notebooklm¶

Interact with NotebookLM via browser automation.

Requires: Browser extension, NotebookLM tab in Gobbler group

gobbler notebooklm list                              # List notebooks
gobbler notebooklm query "What are the key points?"  # Query notebook
gobbler notebooklm info                              # Get notebook info

gobbler-chatgpt¶

Send messages to ChatGPT via browser automation.

Requires: Browser extension, ChatGPT tab in Gobbler group

gobbler chatgpt list                        # List ChatGPT tabs
gobbler chatgpt query "Your message here"   # Send message
gobbler chatgpt last                        # Get last response
gobbler chatgpt history --count 10          # Get history

gobbler-claude¶

Send messages to Claude.ai via browser automation.

Requires: Browser extension, Claude.ai tab in Gobbler group

gobbler claude list                        # List Claude tabs
gobbler claude query "Your message here"   # Send message
gobbler claude last                        # Get last response
gobbler claude history --count 10          # Get history

gobbler-gemini¶

Send messages to Google Gemini via browser automation.

Requires: Browser extension, Gemini tab in Gobbler group, signed into Google

gobbler gemini list                        # List Gemini tabs
gobbler gemini query "Your message here"   # Send message
gobbler gemini last                        # Get last response
gobbler gemini history --count 10          # Get history

Installation¶

Via Claude Code Plugin (Recommended)¶

In Claude Code, add the Gobbler marketplace and install the plugin:

/plugin marketplace add Enablement-Engineering/gobbler
/plugin install gobbler@gobbler-marketplace

Via Git Clone¶

git clone https://github.com/Enablement-Engineering/gobbler.git
cd gobbler
make install

Skills are in skills/gobbler-*/SKILL.md.

Prerequisites¶

• Python 3.11+ and uv package manager
• Docker (for webpage/document conversion): make start-docker
• ffmpeg (for audio): brew install ffmpeg
• Browser extension (for browser skills): See Browser Extension

Backend Services¶

The gobbler CLI connects to these backends:

Start Docker services with make start-docker.