Scrollclaw logo

Scrollclaw

Community
TheMattBerman
scrollclaw

Router and orchestrator for AI UGC video production. Use for broad outcome requests like making a UGC video or running a campaign; it routes step-specific requests to the right sub-skill and can drive the full pipeline from persona research through virality-scored video.

Overview

PublisherTheMattBerman
Repositoryscrollclaw
Skill namescrollclaw
Stars
72
Forks
18
Bundled files
47
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 47 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by TheMattBerman on GitHub. Read the source before you install it.

Installation

Install the Scrollclaw AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/TheMattBerman/scrollclaw/tree/main
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/TheMattBerman/scrollclaw.git \
  .claude/skills/scrollclaw
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Scrollclaw in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Scrollclaw on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Scrollclaw is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

ScrollClaw

AI-generated UGC videos that look like a real person pulled out their phone and started talking. Brands pay $500–$5,000 per UGC video from human creators. This pipeline produces them for $5–$50 in API costs.

The Pipeline

Brand name + URL
      ↓
/brand-setup → One-time brand research + file generation (Step 0)
      ↓
Brand + Audience
      ↓
/persona    → Persona research, creator profiles, approved script
/first-frame → Canonical face image (Nano Banana 2)
/animate    → A-roll talking head clips (Sora 2 i2v)
/b-roll     → Environment + product shots (Kling 3)
/assemble   → Stitch + unified voice + post-production + captions
/score      → Virality gate (70+ to publish)
      ↓
Scroll-stopping UGC video

Core doctrine: Read _system/SKILL.md for format selection, anti-patterns, taste calibration, and pipeline routing rules.

Brand & campaign context: Read _system/references/brand-campaign-context.md for workspace structure, context matrix, and persistence protocol.


Routing

When the user invokes ScrollClaw or asks about UGC video, route to the right sub-skill based on where they are in the pipeline:

User says / wantsRoute to
"Set up a brand" / "Initialize brand" / "New brand" / "Brand research"/brand-setup — one-time, before first campaign
"Make me a UGC video" / "Start a campaign" / "I need ugc"/persona — start at the beginning
"Research this brand" / "Create a creator profile" / "Write a script"/persona
"Generate the first frame" / "Make a face image" / "Nano Banana"/first-frame
"Animate this" / "Make a talking head clip" / "Sora" / "A-roll"/animate
"Generate B-roll" / "Product shot" / "Kling"/b-roll
"Stitch the clips" / "Add captions" / "Post-production" / "Assemble"/assemble
"Score this video" / "Is this ready to publish?"/score
"Run the full pipeline"Start at /persona, proceed sequentially

Broad outcome requests stay at the root. Start the pipeline instead of asking the user to pick a sub-skill.

Stage-specific requests bypass the root and go straight to that sub-skill.

If the user isn't sure where they are, ask: "Where are you in the pipeline?" and show them the 6 steps above.

Contract

Input

  • Required: brand/product context plus a campaign goal, or an existing campaign workspace if resuming
  • Optional: format preference, creator profile, screen recordings/screenshots, existing clips, campaign slug
  • Format: raw text, URL, file paths, or workspace/campaigns/<slug>/ files
  • Source: user prompt, campaign brief, and upstream workspace/brand/ files

Output

  • Produces: either a route decision to the correct stage or a sequential full-pipeline run starting at /persona
  • Format: inline orchestration plus saved workspace artifacts produced by sub-skills
  • Default behavior: broad requests like "make me a UGC video" or "start a campaign" start at /persona and continue stage by stage until blocked by missing inputs, an approval gate, or a dependency failure
  • Downstream use: /persona, /first-frame, /animate, /b-roll, /assemble, and /score

Validation

  • Pre-conditions: workspace exists or can be initialized before generation steps begin
  • Post-conditions: the user knows the current stage, the next stage, and which workspace artifacts were produced
  • Failure checks: do not leave the user at a vague route; if blocked, name the exact missing file, asset, or approval needed

Quick Start

First campaign:

  1. Set up brand context (one-time per brand):
    • Recommended: Run /brand-setup with your brand name + URL — it researches and generates all three brand files automatically
    • Alternative: Copy templates from assets/ and fill in manually (see assets/ for templates)
  2. Initialize workspace:
    bash
    CAMPAIGN="my-campaign"
    mkdir -p workspace/campaigns/$CAMPAIGN/{creators,scripts,frames,clips,scores}
    cp assets/campaign-brief-template.md workspace/campaigns/$CAMPAIGN/brief.md
    touch workspace/campaigns/$CAMPAIGN/output-log.md
    touch workspace/campaigns/$CAMPAIGN/learnings.md
  3. Fill in workspace/campaigns/$CAMPAIGN/brief.md
  4. Run /persona → /first-frame → /animate → /b-roll → /assemble → /score

Check setup first:

bash
bash scripts/check-deps.sh

What ScrollClaw Needs

KeyRequiredUsed by
FAL_KEYYesSora 2 (A-roll) + Kling 3 (B-roll)
REPLICATE_API_TOKENYesNano Banana (first frames)
OPENROUTER_API_KEYRecommendedGemini (virality scoring)
ELEVENLABS_API_KEYOptionalMulti-clip voice consistency (S2S)

Six Formats

FormatDurationBest for
Talking Head15-25sProduct review, honest take
Hook Face + Demo15s maxApp/tool demos
Podcast Clip8-20sAuthority, credibility
Wall of Text4-8sHot take, faceless
Visual Transformation10-25sBefore/after concept
Hybrid Transformation20-30sComplex mechanism explanation

Key Findings (from testing)

  • Sora's native voice > ElevenLabs TTS for talking head. TTS sounds fake.
  • B-roll must be environment-matched. Extract a frame from A-roll → feed to Kling.
  • Captions go LAST — after post-production. Grain degrades caption pills.
  • AI cannot generate realistic app screens. Use real screenshots.
  • ~1 in 3 Sora generations have hand artifacts. Reroll, don't fix the prompt.
  • Multi-frame formats: chain from frame 1. Parallel generation causes face drift.

For full doctrine, format blueprints, anti-patterns, and taste calibration: _system/SKILL.md

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Scrollclaw AI skill do?

Router and orchestrator for AI UGC video production. Use for broad outcome requests like making a UGC video or running a campaign; it routes step-specific requests to the right sub-skill and can drive the full pipeline from persona research through virality-scored video.

Why use Scrollclaw on TypingMind?

Because you install it once and use it with any model. Scrollclaw is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Scrollclaw in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/TheMattBerman/scrollclaw/tree/main. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Scrollclaw?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Scrollclaw?

As many as you like. As long as a model supports skills, you can use Scrollclaw with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Scrollclaw AI skill free?

Yes. It is published on GitHub by TheMattBerman under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇