Session Overview
This Campus VIP working session (61 minutes, March 30, 2026) covers the architecture of modular AI workflows, demonstrating why breaking complex automations into individual skills connected by MCP tools produces more reliable, flexible, and reusable systems than monolithic single-skill approaches. Two complete workflow examples are walked through in detail: the Tutorial Waterfall (5 skills for video-to-content repurposing) and the Opportunity Pipeline (4 skills for partner/student outreach). A participant’s real course-creation workflow is mapped live, and the session includes hands-on plugin customization in Claude.
Date: Sunday, March 30, 2026
Duration: 61 minutes
Participants: James (host), Campus VIP member (professor/educator)
Key Concepts
Modular Skills vs Monolithic Skills
The core architectural principle is that AI workflows should be built from small, focused skills rather than one large skill. Each skill handles a single task and passes its output to the next skill in the chain. This provides debuggability (fix one skill without reworking everything), reusability (same Transcript Analyzer works for YouTube, Zoom, or PDF input), and batch flexibility (5 tutorials but 1 digest email).
The Tutorial Waterfall Workflow
A 5-step workflow that transforms a YouTube video into five pieces of published content: (1) Transcript Analyzer downloads and creates a content brief, (2) Tutorial Body Builder publishes a WordPress draft, (3) Video Announcement Email drafts in FluentCRM, (4) Community Feed Poster creates a discussion in FluentCommunity, (5) Social Media Content publishes to LinkedIn, X, and Facebook via ContentStudio.
The Opportunity Pipeline Workflow
A 4-step workflow for outreach: (1) Scanner uses Apify web scraping to find prospects, (2) Scorer ranks and qualifies them, (3) Pitcher creates personalized outreach emails in FluentCRM, (4) Reporter tracks opens, clicks, and responses.
Org Chart Department Mapping
Every business maps to four AI departments: Sales, Community, Education, and Marketing. A dedicated skill in the Meet Your AI Team course walks you through mapping your specific business to these departments.
Plugin Customization Process
Existing plugins can be duplicated and adapted: open an existing plugin, choose Customize, rename for your use case, modify individual steps, and add/remove skills as needed.
AI Image Generation Alternatives
Three paths: AI Engine + NanoBanana (direct WordPress integration), kie.ai (multi-model service with free tier), and NotebookLM (useful but has spelling errors and no editing).
Step-by-Step: Mapping Your Own Workflow
- Pick one process you repeat
- Write it as a flowchart with every step and decision point
- Identify bottlenecks where YOU are the bottleneck
- Mark what can be delegated — each becomes a potential skill
- Prioritize by impact: saves time or earns money
- Build or adapt existing plugins rather than starting from scratch
- Add meta-workflow steps: announcements, marketing, follow-up
Tools Mentioned
- Claude + Cowork — Workflow orchestration and plugin customization
- FluentCRM — Email automation, member records, central interaction hub
- FluentCommunity — Course hosting, community discussions
- ContentStudio — Social media scheduling
- AI Engine — WordPress AI image generation via NanoBanana
- kie.ai — Multi-model AI connector
- Apify — Web scraping for scanning workflows
- Stripe — Connected to Claude for invoicing
- Canvas LMS — Institutional course delivery
- Modern Event Calendar — WordPress event management
Q&A Highlights
How do skills, agents, and plugins relate? Skills are individual task instructions. Plugins chain multiple skills together. Agents execute the workflow. The skill is the recipe, the plugin is the meal plan, the agent is the chef.
Why not put everything in one skill? Complex single skills break more often, you can’t swap inputs easily, and batch processing requires flexibility that monolithic skills can’t provide.
Can Claude replace NotebookLM for graphics? Yes. AI Engine with NanoBanana and kie.ai both generate images from Claude prompts without NotebookLM’s spelling and editing limitations.
Teachable Moments
“Pick your shots” — Not every workflow needs automation. Filter by: does this save significant time or earn money?
“Meta workflows” — Don’t just automate the core task. Map what happens after: creation, publishing, announcement, promotion, follow-up.
The overwhelm is normal. Pick one workflow, map it, build it, then move to the next.