A walkthrough of the skill file itself — "really quite simple, although it is one of the longer skills" he has written (8:37).
How it opens
- It states the frame up front: the user has asked to be taught something; this is a stateful request; they intend to learn the topic over multiple sessions (
8:37). - Then it describes the shape of the teaching workspace (
8:37).
The philosophy section — defining the AI's vocabulary
He deliberately defines the terms the AI should think in (8:37). Three tiers:
| Tier | Source | How the skill delivers it |
|---|---|---|
| Knowledge | high-quality, high-trust resources (8:37) | highly relevant, interactive lessons |
| Skills | actually doing the thing (8:37) | practice prompts, feedback loops |
| Wisdom | interacting with other learners and practitioners (8:37) | delegation to a community |
Sections of the skill
In order, as shown (9:10):
- shape of the lessons
- the mission /
mission.md - the zone of proximal development
- acquiring knowledge and skills, and how to do that
- acquiring wisdom — how to delegate the learner to a community once they have knowledge and skills
- reference documents
notes.md
Plus a few supporting documents defining formats: learning record format, mission format, resources format, and so on (9:37).
The wisdom rule — the most opinionated instruction
- Written into the skill: when the user asks a question that appears to require wisdom, the default posture is to attempt to answer, but ultimately delegate to a community (
9:10). - The stated intent behind it: not to keep the learner hooked on the agent for everything, but to give them the skills and confidence to go and work with a community — "send you out into the world" (
9:37).
flowchart LR
K["Knowledge<br/>high-trust resources"] --> S["Skills<br/>practice + feedback loops"]
S --> W["Wisdom<br/>community & real-world testing"]
W --> X["Learner leaves the agent"]