SkillAgentSearch skills...

setup

First-time toolkit setup - cloud GPU, file transfer, voice, and prerequisites

Install / Use

npx skills add digitalsamba/claude-code-video-toolkit

Installs into whichever agent you are using.

About this skill
⚡

Claude Commands

Claude Code slash commands

Quality Score

91/100

Category

Operations

Supported Platforms

Claude Code

Our assessment of setup

setup scores 91/100 on our quality scale, 247th of 755 Operations skills we index (top 33%).

Its Claude Commands is 21 KB long, well organised into 51 sections with 27 code examples: a thorough specification that gives an agent plenty to work with.

With 2,544 GitHub stars, it is one of the more widely adopted skills in the catalogue.

Substance
30/30
Structure
20/20
Description
12/15
Adoption
14/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated 5 days ago, so setup is actively maintained.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

setup compared with similar skills

All 4 of these similar skills score higher than setup; compare them before choosing.

SkillScoreStarsUpdatedFormat
setup (this skill)by digitalsamba912.5k5d agoClaude Commands
claude-memby thedotmack100117.9ktodayCLAUDE.md
taste-skillby Leonxlnx100117.6k9d agoCLAUDE.md
Agent-Reachby Panniantong100112.7k9d agoCLAUDE.md
Understand-Anythingby Egonex-AI10093.0k2d agoCLAUDE.md

Frequently asked questions

How do I install setup?
Run npx skills add digitalsamba/claude-code-video-toolkit. The install tabs above show the steps for each supported agent.
Which AI agents does setup work with?
It is written for Claude Code, as a Claude Commands file. Other agents that read the same format can often use it too.
Is setup safe to use?
It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is setup still maintained?
The repository was last updated 5 days ago, so setup is actively maintained.

description: First-time toolkit setup - cloud GPU, file transfer, voice, and prerequisites

Setup

Interactive setup wizard for new users. Checks prerequisites, guides cloud GPU and storage configuration, deploys AI tools, and verifies everything works. Idempotent — safe to run again to add features or verify configuration.

Key Principles

  • Progressive disclosure — one phase at a time, explain before asking
  • Cost transparency — always state what's free and what costs money before any action
  • Idempotent — check what's already configured, skip completed steps
  • Test after each step — don't just save keys, verify they work
  • Skippable — every phase can be skipped, user can return later
  • Claude Code native — use bash tools to check, configure, and verify. This IS the setup wizard.

Entry Point

On invocation, assess current state and adapt:

Step 1: Detect Current State

1. Check .env exists — if not, create from .env.example
2. Read current .env values (which keys are set vs placeholder)
3. Check prerequisites: node --version, uv --version, ffmpeg -version
   (verify_setup.py also reports optional components: Blender)
4. Check Python deps: uv run python -c "import dotenv; import requests"
5. Check Modal CLI: uv run modal --version (if installed)
6. Check for existing Modal apps: uv run modal app list (if authenticated)
7. Summarize what's ready vs what needs setup

Step 2: Present Current State

Show a clear status overview before asking anything:

Toolkit Setup

Prerequisites:
  [check] Node.js 24.x
  [check] Python 3.12
  [check] FFmpeg 7.1
  [check] Python deps installed (uv)

Cloud GPU:      Not configured
File transfer:  Not configured (using free fallback services)
Voice:          Not configured
Music:          Not configured (free key)
ElevenLabs:     Not configured (optional, paid)
3D (Blender):   Not installed (optional)

Let's get you set up! We'll go through each feature one at a time.
You can skip any step and come back later with /setup.

If everything is already configured, show status and offer to verify/test:

Toolkit Setup — All Configured

Cloud GPU:      Modal (8 apps deployed)
File transfer:  Cloudflare R2 (bucket: video-toolkit)
Voice:          Qwen3-TTS (speaker: Ryan)
Music:          ACE-Step API key set
ElevenLabs:     Configured

Everything looks good! Want me to run a quick verification test?

Phase 1: Prerequisites

Check and report. Don't install anything automatically — just tell the user what's needed and how to get it.

Required

  • Node.js 22+ (24 LTS recommended): node --version. If missing: "Install from https://nodejs.org/ — needed to render videos". If it is 18 or 20, it still works, but both are end of life: suggest nvm install 24 && nvm alias default 24
  • Claude Code: Already running if they're seeing this command

Recommended

  • uv: uv --version. If missing: "Install with curl -LsSf https://astral.sh/uv/install.sh | sh (macOS/Linux) or powershell -c \"irm https://astral.sh/uv/install.ps1 | iex\" (Windows) — manages Python and all toolkit dependencies; needed for AI voiceover, image editing, and all cloud GPU tools"
  • Python deps: uv run python -c "import dotenv; import requests". If missing: run uv sync from the toolkit root (uv installs a compatible Python automatically if needed)
  • FFmpeg: ffmpeg -version. If missing: "Install with brew install ffmpeg (macOS) or see https://ffmpeg.org/ — needed for media conversion"

Optional

  • Blender 4.2+: reported by verify_setup.py under "Optional" (or uv run tools/blender_render.py doctor --no-bench). Only tools/blender_render.py uses it (3D character shots, reveals and product shots from a shot spec), and it is experimental. If it's found, say so in the status and that /video will suggest 3D shots where they fit. If it isn't, show one line ("3D (Blender): not installed, optional") and move on: don't count it as missing, and install nothing. docs/blender.md has the install command per platform.

Windows only

Export PYTHONIOENCODING=utf-8 before running any deploy command in this wizard. On Windows, a piped/captured stdout defaults to the cp1252 codec. Modal's deploy progress output contains box-drawing characters it cannot encode, so modal deploy aborts partway through the image build with:

'charmap' codec can't encode characters in position 3-42: character maps to <undefined>

The build itself is fine — only the printing fails, which makes this look like a broken deploy or a network problem rather than an encoding one. Every deploy in Phase 4 fails identically without it.

Set it for the shell you run the deploys in:

export PYTHONIOENCODING=utf-8   # Git Bash
$env:PYTHONIOENCODING = 'utf-8' # PowerShell

Offer to make it permanent so the user does not hit this again outside this wizard:

[Environment]::SetEnvironmentVariable('PYTHONIOENCODING','utf-8','User')

See docs/modal-setup.md for the rest of the Windows prerequisites (notably the Microsoft Store python3 stub).

Output

Prerequisites: 4/4 ready

Moving on to cloud GPU setup...

If anything critical is missing, explain what features require it and ask if user wants to continue or fix first.


Phase 2: Cloud GPU Provider

This is the biggest unlock. Frame it properly.

Explanation (deliver conversationally, not as a wall of text)

Explain what cloud GPU enables:

  • AI voiceovers (Qwen3-TTS) — free, self-hosted text-to-speech
  • AI image generation (FLUX.2) — title backgrounds, thumbnails
  • AI image editing (Qwen-Edit) — transform photos, add effects
  • AI upscaling (RealESRGAN) — enhance image quality
  • AI music generation (ACE-Step): a self-hosted fallback; the default is the free ACE-Step API key in Phase 5
  • Talking heads (SoulX-FlashHead, SadTalker) — animated presenters from a photo
  • Watermark removal (ProPainter) — clean up stock footage
  • AI video clips (LTX-2): a ninth app, deployed on its own after the rest (it needs a Hugging Face token)

Then present the choice:

Option A: Modal (Recommended)

Frame the pitch:

  • $30/month free compute on the Starter plan — just requires adding a payment method
  • A deployed app that isn't called costs nothing. What a call costs, read off Modal's bill for a batch (2026-11-16, docs/modal-setup.md): under a cent a voice line, about a cent an image, about 11 cents an LTX-2 video clip. Don't quote a monthly figure: it depends on what the user makes, and uv run modal billing report shows theirs
  • Faster cold starts than RunPod
  • Scale to zero — no charges when idle
  • Simple deployment: uv run tools/modal_setup.py <app> (deploys and writes .env)

Setup flow:

1. uv sync --extra modal
2. uv run modal setup
   → Opens browser for authentication
   → Creates ~/.modal.toml with credentials
3. Verify: uv run modal app list

Option B: RunPod

Frame honestly:

  • Pay-as-you-go, typically $0.44/hr for GPU time
  • Each job usually costs $0.01-0.50 depending on the tool
  • More GPU variety available
  • --setup flag automates endpoint creation
  • Availability can be spotty — GPUs may not be available in some regions

Setup flow:

1. Create account at https://runpod.io/
2. Get API key from https://runpod.io/console/user/settings
3. Add to .env: RUNPOD_API_KEY=your_key_here
4. Each tool has --setup to create its endpoint automatically

Option C: Both

Explain dual-provider benefits:

  • Switch between providers with --cloud modal or --cloud runpod
  • If one provider is down, use the other
  • Modal for daily use (free tier), RunPod as fallback

Option D: Skip

"No problem — you can still create and render videos with just Node.js. Run /setup anytime to add cloud features later."


Phase 3: Cloudflare R2 (File Transfer)

Only show this phase if user chose a cloud GPU provider.

Explanation

"When you run a cloud GPU tool (like generating a voiceover or editing an image), the tool needs to send your file to the remote GPU and get the result back. Cloudflare R2 is the bridge.

R2 is free — 10GB storage, 10 million operations/month, and zero egress fees. That's more than enough for this toolkit. Without R2, tools fall back to free file hosting services (litterbox, 0x0.st) which are slower and less reliable."

Setup Flow

Guide the user step by step. They'll need to do some of this in their browser:

1. Go to https://dash.cloudflare.com/
   → Create a free account if you don't have one

2. In the sidebar, click "R2 Object Storage"
   → Click "Create bucket"
   → Name it: video-toolkit (or whatever you prefer)
   → Choose automatic region
   → Click "Create bucket"

3. Go to R2 > Overview > "Manage R2 API Tokens"
   → Click "Create API Token"
   → Name: "video-toolkit"
   → Permissions: "Object Read & Write"
   → Scope: Apply to specific bucket → select your bucket
   → Click "Create API Token"
   → COPY the Access Key ID and Secret Access Key (shown only once!)

4. Your Account ID is in the URL: dash.cloudflare.com/{account_id}/r2
   Or find it in the R2 overview sidebar

After user provides the values, write them to .env:

R2_ACCOUNT_ID=xxx
R2_ACCESS_KEY_ID=xxx
R2_SECRET_ACCESS_KEY=xxx
R2_BUCKET_NAME=video-toolkit

Verify

Test R2 connectivity:

uv run python -c "
import sys; sys.path.insert(0, 'tools')
from file_transfer import upload_to_r2, delete_from_r2
import tempfile, os
# Create tiny test file
tmp = tempfile.NamedTemporaryFile(suffix='.txt', delete=False)
tmp.write(b'R2 connectivity test')
tmp.close()
url, key = upload_to_r2(tmp.name, 'setup-test')
os.unlink(tmp.name)
if url:
    print(f'R2 working! Test upload succeeded.')
    delete_from_r2(key)  # Clean up
else:
    print('R2 upload failed — check credentials')
"

Phase 4: Deploy Cloud GPU Tools

Only show this phase if user chose a cloud GPU provider in Phase 2.

Tool Selection

Present the seven choices (eight apps: talking heads is two) with descriptions and let user choose:

Which AI tools would you like to set up?

  1. [recommended] Speech (Qwen3-TTS)
     Free AI voiceovers — 9 speakers, voice cloning, tone presets
     This is the most-used tool in the toolkit.

  2. [recommended] Images (FLUX.2 klein)
     Generate title backgrounds, scene illustrations, thumbnails
     Fast: ~3s per image on warm GPU
     Default klein-4B (Apache-2.0); klein-9B is a non-commercial opt-in

  3. Image Editing (Qwen-Image-Edit-2511)
     Transform photos — add/remove objects, change style, backgrounds
     Note: Large model, ~2 min cold start
     Qwen-Image-2.1 is a non-commercial opt-in (up to 10 input images)

  4. Upscaling (RealESRGAN)
     Enhance image quality 2x or 4x with AI
     Fast: ~5s per image

  5. Music (ACE-Step 1.5)
     Generate background music, jingles, vocals
     8 scene presets: corporate, ambient, dramatic, tension, cta...

  6. Talking Heads (SoulX-FlashHead + SadTalker)
     Animate a portrait photo with lip-sync from audio
     Great for narrator Picture-in-Picture
     SoulX is the default (Modal only, one-off 15GB weights download);
     SadTalker is the cheap draft option

  7. Watermark Removal (ProPainter)
     Remove watermarks from video using AI inpainting

  a. All of the above (recommended)
  s. Skip for now

Recommend "all": there's no cost to having them deployed (they scale to zero), and --all deploys the eight apps behind these seven choices. Say that video clips (LTX-2) are one more command afterwards.

Modal Deployment Flow

Deploy the selected apps with tools/modal_setup.py. It deploys one at a time (Modal rate-limits parallel AppCreate), runs SoulX's populate_weights first, reads each endpoint URL from Modal's API, and writes it to .env. Don't parse URLs out of the deploy log yourself, because Modal wraps and truncates them.

# Windows: requ

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars2.5k
CategoryOperations
Updated5d ago
Forks444

Languages

Python

Trust signals

100/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

No cautions