_scrollclaw-system

Enforce an anti-polish pipeline for generating phone-real UGC videos.

68|15|Updated Mar 24, 2026
One-click install
npx skills add https://github.com/TheMattBerman/scrollclaw --skill scrollclaw-system
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: _scrollclaw-system
Source: https://github.com/TheMattBerman/scrollclaw/tree/main/_system
Command: npx skills add https://github.com/TheMattBerman/scrollclaw --skill scrollclaw-system

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

ScrollClaw prevents AI-generated UGC from looking like polished ads by enforcing an anti-polish pipeline: real persona language, phone-real first frames, handheld energy, authentic audio, and strict post-production order.

Core Features & Use Cases

  • Persona-first scripting: mines real customer language before production so the script sounds like a buyer, not a copywriter.
  • Creator identity locking: keeps the same persistent creator look and voice across clips using canonical first-frame references and reusable creator profiles.
  • First-frame-driven video generation: generates a controlled canonical frame first, then animates via i2v to preserve identity, composition, and color world.
  • Anti-pattern guardrails: blocks common failure modes like random people, text-to-video for established creators, captioning before post-production, and content drift after stitching.
  • Format routing by conversion intent: selects the correct UGC format (Talking Head, Hook Face + Demo, Podcast Clip, Wall of Text, Visual Transformation, Hybrid Transformation) based on the content goal.

Quick Start

Tell the system you are making a UGC video for a specific brand and specify which format goal you want (review, demo, authority, transformation, or faceless hot take), and then follow the pipeline order from brand setup through scoring.

Frequently Asked Questions about _scrollclaw-system

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make AI-generated UGC look like real phone-taken videos?

To make AI-generated UGC look like real phone-taken videos, use an anti-polish pipeline that enforces real persona language, handheld energy, and authentic audio to prevent the content from looking like a polished ad.

What is the best way to keep the same creator identity across multiple AI video clips?

Keeping the same creator identity across multiple AI video clips requires creator identity locking using canonical first-frame references and reusable profiles, preventing identity drift during i2v animation.

How do I select the right UGC video format for my conversion goal?

Selecting the right UGC video format for your conversion goal requires format routing based on content intent, choosing between Talking Head, Hook Face + Demo, Podcast Clip, Wall of Text, or Visual Transformation formats.

Why does my AI UGC content drift after stitching clips together?

AI UGC content drifts after stitching when strict pipeline ordering is ignored, specifically when post-production guardrails are bypassed or text-to-video is used for established creators instead of canonical i2v.

When should I add captions during the AI video production pipeline?

You should add captions during the post-production phase of the AI video production pipeline, strictly after stitching is complete, as captioning before post-production is a documented anti-pattern guardrail.

What is persona-first scripting for UGC videos?

Persona-first scripting for UGC videos mines real customer language before production starts, ensuring the script sounds like an authentic buyer rather than a copywriter, maintaining anti-polish standards.