AgentSkillsCN

start

设置项目并学习如何使用该项目。在首次打开项目时,应先运行此步骤。

SKILL.md
--- frontmatter
name: start
description: Set up the project and learn how to use it. Run this first when opening the project.
user_invocable: true

Short-Form Video Transcriber - Start

When the user runs /start, follow this workflow:

Step 1: Welcome Message

Display this welcome message:


Welcome to the Short-Form Video Transcriber!

This project helps you:

  • Scrape videos from TikTok profiles
  • Transcribe audio to text using AI (Whisper)
  • Create organized summaries by topic

Let me check if your environment is set up...


Step 2: Environment Check

Run these checks:

bash
# Check Python
python3 --version

# Check ffmpeg
which ffmpeg || where ffmpeg

# Check if venv exists
ls -la .venv/bin/activate 2>/dev/null || echo "NO_VENV"

If venv doesn't exist:

Create and set up the environment (this also installs yt-dlp automatically):

bash
python3 -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install -e ".[dev]"
pip install yt-dlp  # Install yt-dlp in the venv

If venv exists:

Activate and ensure yt-dlp is installed:

bash
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install yt-dlp  # Ensure yt-dlp is available
pytest tests/unit/ -v --tb=short 2>&1 | tail -5

If ffmpeg is missing:

ffmpeg is required for audio extraction. Provide platform-specific instructions:

macOS:

bash
brew install ffmpeg

Linux (Debian/Ubuntu):

bash
sudo apt update && sudo apt install ffmpeg

Linux (Fedora/RHEL):

bash
sudo dnf install ffmpeg

Windows:

powershell
winget install ffmpeg

Or download from https://ffmpeg.org/download.html and add to PATH.

After installing ffmpeg, re-run /start to continue setup.

Step 3: Explain Available Commands

After setup is verified, explain the commands:


Setup complete! Here's how to use this project:

Available Commands

CommandWhat it does
/startYou're here! Set up the project and see available commands
/bulkTranscribe ALL videos from a TikTok profile
/transcribeTranscribe specific video URL(s) you paste
/accountsSwitch profiles, add multiple accounts, process several at once
/skillifyTurn a summary into a reusable Claude skill

/bulk - Process Entire Profile

Use this when you want to transcribe all videos from someone's TikTok.

Just run /bulk and I'll ask you for:

  1. The TikTok profile URL (e.g., https://www.tiktok.com/@example_creator)
  2. How many videos to process (all, or a specific number)

I'll then download, transcribe, and summarize each video automatically.

/transcribe - Process Specific Videos

Use this when you have specific video URLs you want to transcribe.

Just run /transcribe and paste the URLs:

code
https://www.tiktok.com/@username/video/123456789
https://www.tiktok.com/@username/video/987654321

You can paste one URL or multiple URLs at once.

/accounts - Manage Multiple Profiles

Want to scrape from different TikTok accounts? Use /accounts to:

  • Switch to a different profile
  • Add multiple profiles to your list
  • Process several profiles at once
  • Remove profiles you no longer need

Your saved profiles are stored in accounts.json.

/skillify - Create Claude Skills from Summaries

Turn any summary into a reusable Claude skill! This command:

  1. Takes a transcript summary you've created
  2. Researches the topic with web search to add depth
  3. Creates a properly formatted Claude skill file
  4. Saves it to your skills directory (global or project-local)

Great for building a knowledge base from video content!

Output

All commands create:

  • transcripts/ - Raw transcripts with metadata
  • summaries/ - Organized summaries grouped by topic
  • summaries/INDEX.md - Master index of all content
  • Skills created via /skillify go to your chosen directory

What would you like to do?

  • Run /bulk to process an entire TikTok profile
  • Run /transcribe to process specific video URLs
  • Run /skillify to turn a summary into a Claude skill
  • Or just paste some TikTok URLs and I'll transcribe them for you!

Step 4: Handle User's Next Action

After explaining, wait for the user to either:

  1. Run /bulk or /transcribe
  2. Paste video URLs directly (treat this as running /transcribe)
  3. Ask questions

If they paste URLs directly without a command, process them as if they ran /transcribe.

Error Handling

If setup fails:

  • Show clear error message
  • Suggest specific fix
  • Offer to retry

If tests fail:

  • Show which tests failed
  • Check if dependencies are installed correctly
  • Suggest reinstalling with pip install -e ".[dev]"