| name | learn-guide |
| description | Generate a personalized learning guide (text + hands-on script) for a technology/topic, tuned to the user's level and case context — like an AI-personalized intro book. Use when the user says "学習ガイド", "入門ガイド", "ハンズオンを作って", "learn guide", "この技術を学びたい", or wants a guided, output-oriented way to learn something. Powers the learning half of learn_cycle.md. |
| triggers | ["学習ガイド","入門ガイド","ハンズオンを作って","learn guide","この技術を学びたい","勉強したい"] |
| user-invocable | true |
| allowed-tools | Read, Write, Bash, Glob, WebFetch, WebSearch |
/learn-guide — Personalized learning guide + hands-on script
Produce a tailored "personalized intro book" for a technology/topic: a structured learning
path plus a runnable hands-on scaffold. The point is output-driven learning — learn by
doing and then publish small notes (高い定着率 + 速い学習速度), per ~/learn_cycle.md L7.
1. Scope the guide (ask if missing)
- Topic: the technology/concept to learn (e.g. "eBPF", "Pulumi", "OpenTelemetry traces").
- Goal / why now: shipping a feature, a case need, general depth — shapes scope.
- Current level: assume the user's background (SRE / インフラ / Go・Ruby・PHP・TS,
Terraform/AWS/Kubernetes) unless told otherwise; calibrate prerequisites accordingly.
- Time budget & depth: quick orientation vs. multi-session deep dive.
2. Ground in primary sources (do not hallucinate)
- Use WebFetch/WebSearch to pull from official docs / source / specs. Prefer 1次情報.
- Verify version-specific facts against current docs; cite every external claim with its URL.
- Flag anything you could not verify as unverified rather than asserting it.
3. Build the guide (Japanese)
Output a structured Markdown guide with these parts (adapt depth to §1):
- 前提知識 — what to already know; quick refreshers/links for gaps
- 全体像 — the mental model in a few sentences + a
[ここに図: 〜] placeholder
- 概念 — core concepts, each defined then contrasted with what the user already knows
- ハンズオン — numbered, runnable steps; each step states the goal, the command/code, and
the expected result, plus the why behind each decision
- 演習 — 2-3 small exercises that force application (not just reading)
- よくある落とし穴 — pitfalls worth knowing (pre-warn only those costing 30+ min)
- 次の一歩 + 1次情報リンク — where to go deeper, with cited official sources
- Japanese style: follow the k16shikano 文章規範 in
skills/write-article/SKILL.md
(one sentence per line, concrete, sparse bold, no LLM-ish phrasing). Plain 常体 is fine for
a study guide; do not force the ぐりもお blog voice here.
4. Hands-on scaffold (runnable)
- Create a working dir (ask for path, else
~/learn/<topic-slug>/) with mkdir -p.
- Write the minimal runnable scaffold the hands-on needs (scripts, config, sample code,
a
README.md that mirrors the ハンズオン steps). Keep commands copy-pasteable.
- Save the guide as
~/learn/<topic-slug>/GUIDE.md (or the user's chosen path).
- Do NOT run install/network commands automatically; show them for the user to run.
5. Close the loop
- Print the guide path and the scaffold location.
- Remind: as you work through it, capture small notes — a fresh, small-granularity
personal-blog post via
/write-article is the 定着装置 in learn_cycle.md. If the
learning came from a case, the generalized version can later become a /company-blog draft.