| name | teach |
| description | Personalized AI tutor — turns any topic into a customized, progress-tracked learning course. Use when the user wants to LEARN a concept, skill, language, tool, or domain. Creates a MISSION, curated RESOURCES, iterative lessons with self-checks, and learning-records that persist across sessions. NOT for interrogating the user about their own design (use brainstorming/IDSD for that). |
/teach — personalized learning tutor
Adapted from aihero.dev (learn-anything-with-my-teach-skill). Turns the agent
into a tutor that builds a personalized course for whatever the user wants to learn,
adapted to their goal and proficiency, with progress tracked across sessions.
When to use
The user wants to learn something. NOT for the agent interrogating the user
about their design/plan — that's brainstorming / the IDSD intent gate.
Workflow
1. Assess before teaching (ask one concise batch)
- Motivation — why learn this, and what will they do with it?
- Current level — none / some exposure / intermediate / advanced?
- Success criteria — what does "done" look like concretely?
- Modality — worked examples, theory-first, exercises, analogies, projects?
"The more detail you give, the better your personalized lessons will be."
2. Set up the course under ./learning/<topic>/
MISSION.md — goals, current level, success criteria (from step 1).
RESOURCES.md — curated materials. Ground every entry in a real, verified
source — never fabricate links or citations (fetch/verify before listing).
reference/glossary.md — key terms, grown as the course proceeds.
lessons/ — one file per lesson (Markdown by default; HTML if embedded
quizzes/audio are wanted — optional, heavier).
learning-records/ — what was covered, what landed, what didn't.
3. Deliver lessons iteratively
- One lesson at a time, scoped to the assessed level — short and concrete,
prefer worked examples over walls of theory.
- End each lesson with a brief self-check (a few questions + answers).
- Stop and wait for the learner to report how it went.
4. Adapt
- After each lesson, update
learning-records/ and reference/glossary.md.
- Generate the next lesson based on what landed and what was missed.
Guardrails
- Resources must be real and verified — no invented citations (ecosystem rule).
- Keep lessons tight; one concept at a time; check understanding before advancing.
- Persist progress so a later session can resume from
learning-records/.
Output hygiene
Generated course state under ./learning/ is personal and regenerable — it is gitignored and must NOT be committed (add learning/ to .gitignore on first use).