NotebookLM is a great tool. But it has nine output options and that’s it. A Claude skills library has no ceiling — and unlike NotebookLM’s fixed menu, your skills are portable across every major AI platform. This is the case for building your own “employees” instead of relying on someone else’s tool list.
The NotebookLM Problem (It’s Not What You Think)
NotebookLM is genuinely useful for what it does: adding sources, chatting with documents, and generating a fixed set of outputs — infographics, slide text, audio overviews, video summaries, and a few others. For a specific job, it’s excellent.
The problem is the ceiling. Nine outputs. Configured by Google. Not by you.
If you want an infographic that looks a specific way, a script structured around your teaching style, a community post that uses your voice, or a checklist that maps to your course framework — NotebookLM can’t do that. You’d need a different tool for each one. And then you’d need to learn each tool. And then you’d need to manage the output manually.
What 214 Skills Actually Looks Like
James’s skills library — free to TrainingSites community members — has 214 agent skills organized into 7 categories built specifically for education businesses:
- Performance & Growth — tracking outcomes, student results, community health
- Marketing & Sales — offers, emails, social copy, launch sequences
- Content Production — tutorials, scripts, repurposing, YouTube workflow
- Live Learning — class prep, cohort design, flipped session formats
- Automation & Integration — MCP connections, workflow orchestration
- Research & Strategy — competitor analysis, ICP research, positioning
- Content Authority — topical authority mapping, library organization
Each skill is a trained employee with a specific job. You don’t learn how to do the task — you task the employee and review the output.
The Key Difference: Skills Are Cross-Platform
Agent skills follow an open standard. That means the 214 skills in the TrainingSites library work in:
- Claude (chat interface)
- Claude Cowork (agentic desktop mode)
- Claude Code
- Gemini
- OpenAI / ChatGPT
- Manus and other major AI platforms
These are transferable employees. You build them once. They work wherever you work.
Live Demo: Chat to Community Post
In the video, James runs a live demo using the “Chat to Educational Post” skill. He has a conversation with Claude about a video he’s making — back and forth, refining ideas. Then he activates one skill with a short instruction: create a 150-word engagement post with hook and discussion prompts, and post it to the community.
The skill extracts the teaching moments from the conversation, finds the right community space, writes the post, and publishes it — without James touching a keyboard again. He describes it as being on probation: “I’m spending time with them and saying, hey, is it doing stuff the way I want?”
This is the operating model. You define the outcome. The employee figures out the execution.
The Education Industry Is Splitting in Two
This is the part worth sitting with:
“The education industry is splitting in two. Content creators are the people losing. Transformation facilitators are the people winning. And the reason is because skills handle all of that other work. Now you just get to spend time with people.”
If your business model depends on producing and selling content, the tools your students have access to are eroding that moat every week. If your model depends on live facilitation, accountability, and outcomes — AI handles the operational layer while you do the work only you can do.
Where to Start
Download a few skills from the library. Filter by category or keyword. Install one in Claude. Give it a task. See what comes back. Adjust if needed — that’s what “probation” looks like.
The full skills library (214 skills, free to members) is at trainingsites.io/join.
