Cal AI Co-Founder Shares Full AI UGC Workflow and Claude Skill for App Growth

Jake Castillo, co-founder of the calorie-tracking app Cal AI—which surpassed 15 million downloads and was acquired by MyFitnessPal—has shared his complete operational workflow for scaling mobile apps using AI-generated User-Generated Content (AI UGC). Shared via AI creator Nico (@nicos_ai), the release includes a structured pipeline that avoids random video generation, 13 prompts tailored for leading video generation models, and a specialized Claude skill.

Image source: Nico (@nicos_ai) / Jake Castillo (@jakecastilloooo)

Rather than relying entirely on expensive paid acquisition campaigns, Cal AI scaled to a $50M revenue run rate by building a high-volume distribution engine powered by over 300 influencer retainers and 20+ dedicated social accounts. The newly published system adapts that proven testing and distribution strategy for modern AI video generation workflows.

1. Finding Proven Viral Formats in Your Niche

The foundation of the workflow begins by identifying formats and narrative structures that are already generating strong organic traction within a specific niche, rather than generating unguided creative concepts from scratch.

• Format Reverse Engineering: Marketers study top-performing accounts within their target category (e.g., health and fitness, productivity, personal finance) to dissect hooks, camera angles, pacing, and retention triggers.
• Frictionless Product Placement: The system pinpoints native video structures where core app actions (such as scanning a meal before eating) integrate seamlessly without disrupting the natural flow of the creator's content.

2. Testing One AI Video Before Batch Scaling

After isolating a promising format, the playbook advises against committing large budgets or commissioning heavy production batches prematurely. Instead, teams validate conversion hypotheses with rapid AI prototyping.

• Single AI Prototype Generation: Teams generate exactly one AI UGC video using modern video generation models to benchmark the concept.
• Data-Driven Iteration: Early viewer retention and engagement metrics determine whether the angle has traction before any additional resources are invested.
• Handoff to Real Creators: Winning concepts validated through AI tests are handed off to human creators and influencers to scale up organic social reach and performance ad campaigns.

3. 13 Prompts and Multi-Model Claude Skill

To streamline execution across different video generation architectures, Jake Castillo packaged the system with 13 structured prompts and a dedicated Claude skill.

• Support for Top Video Generation Models: Prompt instructions are tailored to account for model-specific prompt comprehension styles and camera motion parameters.
• Claude Skill Automation: By inputting an app's unique value proposition and target audience characteristics into the Claude skill, creators can instantly generate production-ready scripts and detailed scene-direction prompts aligned with the 13 proven formats.

Original source

• Nico on X (@nicos_ai): Cal AI Co-Founder's AI UGC Workflow Breakdown
• Jake Castillo on X (@jakecastilloooo): AI UGC 13 Prompts and Claude Skill Announcement