スキル一覧に戻る

learn-guidance

pjt222
更新日 2 days ago
5 閲覧
17
2
17
GitHubで表示
メタaitestingdesign

について

このClaude Skillは、開発者が新しい技術やスキルを体系的に学ぶためのAI学習コーチとして機能します。既存の知識を評価し、個人に合わせた学習パスを設計し、教材を段階的に解説し、適応型難易度のテストで理解度を確認します。新しい技術を始めるとき、ドキュメントの多さに圧倒されるとき、記憶定着のための間隔反復が必要なとき、あるいは分野を跨ぐ際のギャップ分析が必要なときにご利用ください。

クイックインストール

Claude Code

推奨
メイン
npx skills add pjt222/agent-almanac -a claude-code
プラグインコマンド代替
/plugin add https://github.com/pjt222/agent-almanac
Git クローン代替
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/learn-guidance

このコマンドをClaude Codeにコピー&ペーストしてスキルをインストールします

ドキュメント

Learn (Guidance)

Guide person thru structured learning for new topic / tech / skill. AI coaches — assess starting knowledge, plan path, walk at right pace, test understanding w/ qs, adapt on feedback, consolidate for retention.

Use When

  • Person wants new tech / framework / lang / concept, doesn't know where to start
  • Overwhelmed by docs / resources, needs structured path
  • Keeps forgetting, needs spaced repetition guidance
  • Transitioning domains (e.g., backend → frontend), needs gap analysis
  • Wants accountability + structure for self-directed
  • After meditate-guidance clears noise → space for focused learning

In

  • Req: What to learn (topic, tech, skill, concept)
  • Req: Purpose (job, interest, project, career change)
  • Opt: Current knowledge level (self-assessed / demonstrated)
  • Opt: Time avail (hr/day/week, deadline)
  • Opt: Preferred style (reading, hands-on, video, discussion)
  • Opt: Prior failed attempts (what didn't work)

Do

Step 1: Assess — Starting Position

Before designing path → understand where person stands.

  1. Ask experience: "What do you already know about X?"
  2. Ask adjacent knowledge: "What related topics familiar?" (become bridges)
  3. Some knowledge claimed → calibration q revealing depth vs surface
  4. Note vocab: domain terms correct, approximate, or none?
  5. ID goal specifically: "After learning this, what want to do?"
  6. ID motivation: curiosity, practical, career, creative
Starting Position Assessment:
┌───────────────┬────────────────────────────┬──────────────────────────┐
│ Level Found   │ Indicators                 │ Path Approach            │
├───────────────┼────────────────────────────┼──────────────────────────┤
│ No exposure   │ No vocabulary, no mental   │ Start with "what" and    │
│               │ model, everything is new   │ "why" before "how"       │
├───────────────┼────────────────────────────┼──────────────────────────┤
│ Surface       │ Has heard terms, no hands- │ Fill vocabulary gaps,    │
│ awareness     │ on experience, vague model │ then move to hands-on    │
├───────────────┼────────────────────────────┼──────────────────────────┤
│ Partial       │ Some experience, gaps in   │ Identify specific gaps   │
│ knowledge     │ understanding, can do some │ and target them directly │
│               │ things but not others      │                          │
├───────────────┼────────────────────────────┼──────────────────────────┤
│ Refresher     │ Knew it before, now rusty  │ Quick review + practice  │
│ needed        │                            │ to reactivate knowledge  │
└───────────────┴────────────────────────────┴──────────────────────────┘

→ Clear picture of starting pos, goal, constraints. Warm + encouraging, not exam — frame qs as curiosity.

If err: Can't articulate current level → ask to describe recent attempt. Concrete stories reveal level > self-assessment. Embarrassed → normalize: "Everyone starts somewhere — knowing where helps design best path."

Step 2: Plan — Design Path

Create structured path from current pos → goal.

  1. Break into 4-7 milestones (not too granular, not vague)
  2. Order by dependency: what before what?
  3. Per milestone → core concept + core skill
  4. Estimate time per milestone from avail hours
  5. ID first milestone — learning begins
  6. Build early wins: first milestone achievable quickly → momentum
  7. Present visually: numbered list w/ brief descriptions

→ Path person sees + understands. Feels manageable, not overwhelming. Person can point to any milestone + understand why there.

If err: Too long → goal ambitious for time → discuss scope reduction. Too short → topic simpler than expected / milestones too coarse, decompose.

Step 3: Guide — Walk Material

Per milestone → guide at right pace.

  1. Brief overview: "In this section, learn X → do Y"
  2. Present in small chunks — one concept per chunk
  3. Use preferred style: reading → text; hands-on → exercises; discussion → Socratic
  4. Connect each new concept to something known (from assessment)
  5. Concrete examples before abstract definitions
  6. Using docs → guide thru relevant sections, don't send off to read alone
  7. Pause per chunk: "Make sense so far?"

→ Person progresses w/ comprehension, not just exposure. Can explain each concept in own words before next. Pace feels right.

If err: Struggling → slow down, check missing prereqs. Breezing → speed up, don't waste time. Material confusing (bad docs) → clearer explanation + note resource quality.

Step 4: Test — Check Understanding

Verify learning w/ application qs, not recall.

  1. Prediction: "What would happen if changed X?"
  2. Comparison: "How diff from Y learned earlier?"
  3. Application: "How use this to solve Z?"
  4. Debug: "This code has bug related to what we learned — spot it?"
  5. Celebrate correct answers specifically: "Yes — reason that works is..."
  6. Incorrect → explore reasoning: "Walk me thru your thinking"
  7. Never frame incorrect as failure → diagnostic info

→ Testing reveals working model vs surface recall. Working handles variations; surface can't. Collaborative, not exam.

If err: Can't answer application → learning too passive, need hands-on before more. Recall yes, application no → concepts individual but not integrated → focus connections.

Step 5: Adapt — Adjust Path

Based on tests + feedback → adjust.

  1. Easy milestone → combine w/ next, or deepen content
  2. Hard → break smaller, add prereq review
  3. Interest shifts → adjust to curiosity where possible — engagement drives retention
  4. Fatigued → break + review later vs push thru
  5. Teaching approach not working → diff modality (reading→doing, abstract→concrete)
  6. Update path + communicate: "Based on how this went, suggest adjust..."

→ Path evolves on real data. No fixed curriculum survives contact w/ actual learner — adaptation = value.

If err: Repeated adaptations still struggle → fundamental prereq gap missed in assessment → return Step 1, probe deeper. Losing motivation → discuss original goal — adjusting goal sometimes > changing path.

Step 6: Review — Consolidate + Plan Next

Solidify + setup continued learning.

  1. Summarize covered: "Today learned X, Y, Z"
  2. Ask person state key takeaway in own words
  3. Brief practice for indep work (not homework — optional reinforcement)
  4. Recommend 2-3 resources (docs, tutorials, examples)
  5. Spaced repetition → schedule reviews: "Review again in 2 days, then week"
  6. Next milestone: "Next time, tackle..."
  7. Ask feedback: "What worked? What diff?"

→ Person leaves w/ clear understanding of learned, practice, next. Clean closing, not abrupt stop.

If err: Can't state takeaway → covered too much / too little stuck. ID one concept most needing reinforcement + focus review. No motivation for indep practice → path more self-contained (all learning in sessions).

Check

  • Starting position assessed before path designed
  • Path has clear milestones ordered by dependency
  • Material presented in small chunks w/ comprehension checks
  • Testing used application qs, not just recall
  • Path adapted ≥1 based on actual progress
  • Session ended w/ summary, practice suggestion, next steps
  • Person felt encouraged, not tested / judged

Traps

  • Info dumping: All material at once vs pacing thru milestones. Overwhelm kills learning.
  • Skip assessment: Assume level vs check. Frontend expert learning backend may know adjacent concepts but not ones you expect.
  • Teach to average: Person faster / slower than expected → pace must change. Stick to plan despite feedback → wastes time / loses them.
  • All theory, no practice: Understanding requires doing. Every milestone → practice element.
  • Ignore motivation: Person doesn't see why matters → won't retain. Connect concepts to stated goal.
  • Overload sessions: Too much one sitting. Less w/ retention > more w/ forgetting.
  • Coach-as-lecturer: Coach guides exploration, doesn't monologue. More qs than answers.

  • learn — AI self-directed variant → systematic knowledge acquisition
  • teach-guidance — coach person to teach others; complementary to learning coaching
  • meditate-guidance — clear noise before learning → improves focus + retention
  • remote-viewing-guidance — shares structured observation supporting learning from experience

GitHub リポジトリ

pjt222/agent-almanac
パス: i18n/caveman-ultra/skills/learn-guidance
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

関連スキル

content-collections

メタ

このスキルは、Content Collections(Markdown/MDXファイルを型安全なデータコレクションに変換するTypeScriptファーストのツール)の本番環境でテストされた設定を提供します。Zodバリデーションによる型安全性を実現し、ブログ、ドキュメントサイト、コンテンツ重視のVite + Reactアプリケーション構築時にご利用ください。Viteプラグインの設定、MDXコンパイルから、デプロイ最適化、スキーマバリデーションまで、すべてを網羅しています。

スキルを見る

polymarket

メタ

このスキルは、開発者がPolymarket予測市場プラットフォームを活用したアプリケーション構築を可能にします。API統合による取引や市場データの取得に加え、WebSocketを介したリアルタイムデータストリーミングにより、ライブ取引や市場活動を監視できます。取引戦略の実装や、ライブ市場更新を処理するツールの作成にご利用ください。

スキルを見る

creating-opencode-plugins

メタ

このスキルは、開発者がコマンド、ファイル、LSP操作など25種類以上のイベントタイプにフックするOpenCodeプラグインを作成することを支援します。JavaScript/TypeScriptモジュール向けに、プラグイン構造、イベントAPI仕様、および実装パターンを提供します。カスタムイベント駆動ロジックでOpenCode AIアシスタントのライフサイクルをインターセプト、監視、または拡張する必要がある場合にご利用ください。

スキルを見る

sglang

メタ

SGLangは、高性能なLLMサービングフレームワークであり、RadixAttentionプレフィックスキャッシュを活用したJSON、正規表現、エージェントワークフロー向けの高速で構造化された生成を特長とします。特にプレフィックスが繰り返されるタスクにおいて、大幅に高速な推論を実現し、複雑な構造化出力やマルチターン対話に最適です。制約付きデコードが必要な場合や、広範なプレフィックス共有を伴うアプリケーションを構築する場合は、vLLMなどの代替案ではなくSGLangを選択してください。

スキルを見る