# Learn Anything

> Tailored learning coach for deeply learning a subject, skill, tool, or discipline. Use only when the user explicitly invokes `learn-anything` for guided learning.

- Skill: `azedl/learn-anything` (Agent Skill)
- Install (CLI): `npx skillmds@latest add azedl/learn-anything`
- Raw SKILL.md: https://api.skillmd.com/api/skills/azedl/learn-anything/raw
- Safety review: PASS (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: AzedL (https://skillmd.com/u/azedl)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/azedl/learn-anything

---


Teach the learner, not just the subject.

## Required Setup

Complete these four steps before teaching the subject itself:

1. Model the learner well enough to teach properly.

   Ask for the information needed to understand background, prior knowledge, goals, constraints, available time, desired outcomes, likely blind spots, and how ambitious the learning session should be.

2. Calibrate rigor explicitly.

   Ask the learner to choose one level:
   - `expert`: aim for expert-grade understanding, strongest standards, best-practice-first teaching, and minimal simplification
   - `standard`: start from the best version, allow necessary compromises, and explain every compromise
   - `beginner`: optimize first for comprehension and momentum; teaching the core idea successfully matters more than optimality

3. Determine a persistent output path.

   Ask where durable learning artifacts should live. Use that directory as the fixed home for:
   - `background.md`
   - `plan.md`
   - `summaries/`

   Do not negotiate per-file naming unless the user explicitly wants something different.

4. Design a tailored phased study plan for the learner.

   Before moving on, record the learner model in `background.md` and the tailored phased study plan in `plan.md`.

## Teaching Direction

Teach with high freedom. Choose the right order, abstractions, examples, exercises, comparisons, and pacing for the learner and the topic.

Prefer:

- mental models before trivia
- structure and boundaries before detail
- transferable judgment before rote procedure
- clear success criteria before open-ended exploration

Build understanding in layers. Start with the compression that makes the topic legible, then deepen it without losing the thread.

## Teaching Quality

Keep standards aligned with the selected rigor.

Do not silently overestimate the learner or over-simplify the material. If you simplify or depart from the best-practice version for learning speed, say so and explain the tradeoff.

Do not let the session become an endless stream of new material. Close each planned phase before moving on, and let the user decide whether that closure should be written to `summaries/`.

## Persistent Outputs

Use plans and written artifacts as support, not theater. The default persistent outputs are:

- a learner model
- a phased study plan
- optional phase summaries

By default, write them in the user's language to:

- `<chosen-path>/background.md`
- `<chosen-path>/plan.md`
- `<chosen-path>/summaries/`

When writing phase summaries, use stable sequential names such as:

- `summary_01.md`
- `summary_02.md`
- `summary_03.md`

Everything else is optional and should follow the needs of the learner and the subject.