teach-guidance
について
このスキルは、プレゼンテーション、ドキュメンテーション、メンタリングなどの技術的コンテキストにおいて、効果的な指導や説明技術を開発者に指導します。コンテンツの構成、異なる聴衆の専門知識レベルへの対応、ソクラテス的問いかけなどの手法を用いた明確さの向上を支援します。講演の準備、チュートリアルの作成、同僚のメンタリング時に複雑なトピックを効果的に伝えるためにご活用ください。
クイックインストール
Claude Code
推奨npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/teach-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
Teach (Guidance)
Guide person → effective teacher/explainer/presenter. AI = teaching coach: assess what to communicate to whom, structure for clarity, rehearse, refine on feedback, support delivery, reflect.
Use When
- Person needs to present tech content + wants prep
- Wants better docs, tutorials, explanations
- Struggles to explain across expertise levels
- Mentoring colleague | junior dev
- Prepping for talk, workshop, knowledge-sharing
- After
learn-guidanceacquired knowledge → now transfer
In
- Required: What teach/explain (topic, concept, system, process)
- Required: Audience (expertise, context, relationship)
- Optional: Format (presentation, doc, 1:1 mentoring, workshop)
- Optional: Time constraints (5m explanation, 30m talk, written)
- Optional: Prev attempts + what didn't work
- Optional: Person's comfort w/ topic (deep expert vs recent learner)
Do
Step 1: Assess — Teaching Challenge
Before structuring, understand full context.
- Ask what teach + why: "What concept needs to land, what if not?"
- ID audience: "Who explaining to? What know already?"
- Assess person's understanding: deep enough to teach? Else suggest
learn-guidancefirst. - ID format: presentation, doc, conversation, code review, pair prog
- Success criteria: "How know audience understood?"
- Surface fears: "What part most nervous?"
Teaching Challenge Matrix:
┌──────────────────┬──────────────────────────┬──────────────────────────┐
│ Challenge Type │ Indicators │ Focus Area │
├──────────────────┼──────────────────────────┼──────────────────────────┤
│ Knowledge gap │ "I sort of know it │ Deepen their own under- │
│ │ but can't explain it" │ standing first (learn) │
├──────────────────┼──────────────────────────┼──────────────────────────┤
│ Audience gap │ "I don't know what │ Build audience empathy │
│ │ they already know" │ and calibration │
├──────────────────┼──────────────────────────┼──────────────────────────┤
│ Structure gap │ "I know it all but │ Organize content into │
│ │ don't know where to │ a narrative arc │
│ │ start" │ │
├──────────────────┼──────────────────────────┼──────────────────────────┤
│ Confidence gap │ "What if they ask │ Practice and preparation │
│ │ something I can't │ for edge cases │
│ │ answer?" │ │
└──────────────────┴──────────────────────────┴──────────────────────────┘
Got: Clear picture: what, to whom, what format, what constraints, where least confident.
If err: Can't articulate audience → create persona: "Imagine 1 specific person. What know? Care about?" Can't articulate topic → may need to learn deeper first.
Step 2: Structure — Clarity
Help build clear narrative for explanation.
- ID single core msg: "If audience remembers 1 thing, what?"
- Build outward: what context needed before, what details after?
- Inverted pyramid: most important first, supporting after
- Tech content patterns:
- Concept: What → Why → How → Example → Edge cases
- Tutorial: Goal → Prereqs → Steps → Verification → Next steps
- Architecture: Problem → Constraints → Solution → Tradeoffs → Alternatives considered
- Debugging: Symptom → Investigation → Root cause → Fix → Prevention
- Each section clear purpose: doesn't serve core msg → cut
- Plan transitions: "Covered X. Building on that, need to understand Y because..."
Got: Outline where every element serves core msg. Logical + inevitable — each section naturally → next.
If err: Structure keeps growing → scope too broad, cut. Flat (everything same level) → hierarchy needs work, ID primary vs supporting. Resists structure ("just explain naturally") → natural works for simple, fails for complex; structure = scaffold.
Step 3: Practice — Rehearse
Person practices explaining, AI = audience.
- Ask explain as to actual audience
- Listen w/o interrupt first pass — find natural flow
- Note where clear vs confused/vague
- Note jargon audience may not know
- Note skipped steps or assumed knowledge
- Note too long on easy parts, rush hard parts
- Time if constraint
Got: First-draft revealing natural patterns — strengths to build on, habits to adjust. Low-stakes: "Rough draft, not performance."
If err: Freezes/says "don't know where to start" → back to Step 2 structure, explain 1 section at time. Self-critical ("terrible") → redirect specifics: "X very clear — let's match Y to that quality."
Step 4: Refine — Improve from Feedback
Specific, actionable feedback.
- Lead w/ strengths: "X using Y analogy was effective because..."
- ID biggest improvement opp (not all, focus on 1-2)
- Specific alternatives: "Instead of [complex], try [simpler]"
- Curse of knowledge: places expertise → skip steps audience needs?
- Audience calibration: depth right? too shallow/deep?
- Analogies accurate? (Misleading > no analogy)
- Re-explain refined section → test improvement
Got: Targeted feedback measurably improves. Difference between 1st + 2nd attempt felt. Constructive — what to do, not avoid.
If err: Defensive about feedback → reframe "this was unclear" → "audience might not follow here, how clearer?" Refined no better → may be structural (Step 2), back to outline.
Step 5: Deliver — Support During
Live → support during.
- Live: prep answers to likely Q's in advance
- Docs: review written for clarity, structure, audience calibration
- Prep "I don't know" moment: "If asked something can't answer, say: 'Great Q — I'll look into it + follow up.' Always acceptable."
- Encourage interaction: prep check Q's for audience
- Recovery plans: audience lost, bored, ahead?
- Coaching during: brief specific prompts ("slow down", "they look confused — check in")
Got: Person feels prepped + supported. Has answers for likely Q's, strategies for unexpected, confidence not knowing everything OK.
If err: Anxiety primary blocker → address direct: prep reduces anxiety, acknowledging nervousness creates connection. Format keeps changing → accept format + adapt vs control conditions.
Step 6: Reflect — Analyze What Worked
Post-event, guide reflection.
- "What went well? Proud of?"
- "Where audience most engaged? Least?"
- "Anything surprise about audience response?"
- "If could change 1 thing, what?"
- Connect reflection to principles: "Part that worked used [tech]. Apply more broadly."
- ID 1 specific improvement goal next time
- Celebrate accomplishment: teaching = skill improving w/ practice
Got: Concrete insight, not vague feelings. 1 actionable improvement next time.
If err: Only sees negatives → redirect specific moments worked. Only positives → probe areas audience confused. No reflection (moves on) → reflection = where most durable improvement happens, even 5 min matters.
Check
- Challenge assessed before structuring (audience, format, constraints)
- Core msg ID'd, structure organized around it
- Practiced ≥1× before delivery
- Feedback specific, actionable, measurable improvement
- Prepared for Q's, uncertainty, audience adaptation
- Post-delivery reflection ID'd ≥1 specific improvement
- Coaching encouraging throughout — teaching hard, acknowledge
Traps
- Coach content vs teaching: Helping learn material vs present. Need to learn → use
learn-guidancefirst. - Over-structuring: Rigid structure → person's natural voice lost. Structure supports style, not replace.
- Perfectionism trap: Rehearsing endless vs delivering. Diminishing returns — push to delivery.
- Ignore audience diversity: Mixed audience → layered explanation: core for all, details for experts, analogies for newcomers.
- Feedback overload: Too many notes → overwhelms. Focus 1-2 highest impact.
- Neglect emotional prep: Teaching anxiety real. Confidence as important as content.
→
teach— AI self-directed variant for calibrated knowledge transferlearn-guidance— coaching person through learning; prereq to teaching effectivelylisten-guidance— active listening helps teachers respond to audience real-timemeditate-guidance— calm anxiety + focus before teaching event
GitHub リポジトリ
関連スキル
executing-plans
デザインexecuting-plansスキルは、完全な実装計画があり、それを管理されたバッチでレビューチェックポイントを設けながら実行する場合に使用します。このスキルは計画を読み込んで批判的にレビューした後、小さなバッチ(デフォルトは3タスク)でタスクを実行し、各バッチの間に進捗状況を報告してアーキテクトのレビューを受けます。これにより、品質管理チェックポイントが組み込まれた体系的な実装が保証されます。
requesting-code-review
デザインこのスキルは、コードレビュアーサブエージェントを起動し、処理を進める前に要件に対してコード変更を分析します。タスク完了後、主要な機能の実装後、またはmainブランチへのマージ前などに使用すべきです。このレビューは、現在の実装と元の計画を比較することで、問題を早期に発見するのに役立ちます。
connect-mcp-server
デザインこのスキルは、開発者がHTTP、stdio、またはSSEトランスポートを使用してMCPサーバーをClaude Codeに接続するための包括的なガイドを提供します。GitHub、Notion、カスタムAPIなどの外部サービスを統合するためのインストール、設定、認証、セキュリティについて解説しています。MCP統合のセットアップ、外部ツールの設定、またはClaudeのModel Context Protocolを扱う際にご利用ください。
web-cli-teleport
デザインこのスキルは、タスク分析に基づいて開発者がClaude Code WebとCLIインターフェースの選択を支援し、これらの環境間でのシームレスなセッションテレポーテーションを可能にします。Web、CLI、モバイル環境を切り替える際のセッション状態とコンテキストを管理することで、ワークフローを最適化します。様々な段階で異なるツールを必要とする複雑なプロジェクトにご活用ください。
