スキル一覧に戻る

listen-guidance

pjt222
更新日 Yesterday
5 閲覧
17
2
17
GitHubで表示
デザインaidesign

について

`listen-guidance`スキルは、受容的な心構えの育成、反射的な言い換えの実践、明確化の質問の使用など、積極的傾聴の技法をユーザーに指導します。このスキルは、ユーザーがコミュニケーションの改善、困難な会話への準備、頻繁な誤解の解消を必要とする際に、開発者が統合できるよう設計されています。本スキルは、これらの手法を日常のコミュニケーションに実践し統合するための指導に重点を置いています。

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/listen-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

Listen (Guidance)

Guide person in developing + practicing active listening. AI coaches comm — prepare receptive mindset, practice full-presence attending, develop reflective paraphrasing, use clarifying qs effectively, synthesize understanding, integrate into daily comm.

Use When

  • Improve comm in relationships, teams, professional settings
  • Keeps misunderstanding / feels misunderstood
  • Prep difficult conversation (feedback, conflict, negotiation)
  • Leadership role wants to listen more effectively to team
  • Talks > listens, wants to change pattern
  • After meditate-guidance cultivates presence → direct toward others

In

  • Req: Goal (improve general comm, prep specific conversation, pro skill)
  • Opt: Ctx (workplace, personal relationship, team dynamics)
  • Opt: Known challenges (interrupt, mind wandering, emotional reactivity, advice-giving)
  • Opt: Specific conversation prepping
  • Opt: Feedback received about listening

Do

Step 1: Prepare — Receptive Mindset

Before skills → help understand + enter receptive state.

  1. Hear vs listen: "Hearing passive — sounds enter. Listening active — receive, process, understand"
  2. ID their habits: "When someone talks, what is mind usually doing?"
  3. Common patterns to surface:
    • Planning response while other still talking
    • Judging (agree/disagree) vs understanding
    • Problem-solving — jump to solutions before speaker finished
    • Relating — connecting everything to own experience
    • Filtering — hearing only parts that interest / confirm view
  4. Set intention: "For this session, will focus on..."
  5. Brief centering: 3 slow breaths, release agenda of needing to say something

→ Person ID'd ≥1 habitual pattern to change + clear intention for practice. Calm + present, not performance-anxious.

If err: Can't ID pattern → may be unconscious → suggest notice body when someone talking (tension, restlessness, urge to speak). Self-conscious → normalize: "Everyone has these — noticing is first step to choice."

Step 2: Attend — Full Presence

Guide thru giving full attention to speaker.

  1. Physical: eye contact (comfortable, not staring), body orientation, open posture, stillness
  2. Mental: suspend internal monologue, notice drift, gently return
  3. Setup exercise: "I speak 2 min. Your job: only listen — no responding, note-taking, just receiving"
  4. AI speaks moderately complex topic 2 min
  5. After ask: "What noticed about attention? Where went?"
  6. Repeat if helpful, topic triggering ID'd pattern (something they'd want to fix, judge, relate to)

→ Person experiences diff between habitual + intentional listening. Notices drift + practices return. Even brief full presence valuable.

If err: "Was listening whole time" → ask specific content qs — inability to recall = inattention felt like attention. Can't stop internal monologue → focus on speaker's breath rhythm / pace as anchor — occupies analytical mind while keeping attention on speaker.

Step 3: Reflect — Paraphrasing

Teach mirror understanding back.

  1. Purpose: "Paraphrasing shows speaker heard + lets correct misunderstandings"
  2. Form: "What I hear you saying is..." / "Sounds like..." / "If I understand correctly..."
  3. Emphasize: paraphrase meaning not words — own words to show understood idea
  4. Practice: AI makes statement, person paraphrases
  5. Simple: factual w/ clear content
  6. Complex: emotional content, mixed signals, implicit meaning
  7. Feedback per: "Captured main idea. Might include feeling behind..."
Paraphrase Quality Ladder:
┌──────────────┬────────────────────────────────────────────────────────┐
│ Level        │ Example                                                │
├──────────────┼────────────────────────────────────────────────────────┤
│ Parrot       │ "You said you're frustrated with the project"          │
│ (repeating)  │ → Too literal, doesn't show understanding              │
├──────────────┼────────────────────────────────────────────────────────┤
│ Content      │ "The project isn't going the way you expected"         │
│ (facts)      │ → Captures meaning, misses feeling                     │
├──────────────┼────────────────────────────────────────────────────────┤
│ Feeling      │ "You're feeling stuck because the project keeps        │
│ (emotion)    │ hitting obstacles"                                      │
│              │ → Captures both content and emotion                     │
├──────────────┼────────────────────────────────────────────────────────┤
│ Meaning      │ "This matters to you because you put real effort in,   │
│ (full)       │ and the obstacles feel like they're dismissing that"    │
│              │ → Captures content, emotion, and deeper meaning         │
└──────────────┴────────────────────────────────────────────────────────┘

→ Person paraphrases content consistently + reaches feeling ≥1. Experiences how paraphrasing changes dynamic — "speaker" (AI) feels heard.

If err: Too literal (parrot) → encourage: "Try completely diff words w/ same meaning." Jumps to interpretation vs reflection → redirect: "Before interpreting, mirror what said." Awkward → acknowledge unnatural at first, natural w/ practice.

Step 4: Explore — Clarifying Qs

Teach deepen understanding thru placed qs.

  1. Distinguish types:
    • Open: "What was that like for you?" (invites exploration)
    • Clarifying: "When you say X, do you mean A or B?" (resolves ambiguity)
    • Deepening: "What about that matters most to you?" (goes below surface)
    • Leading: "Don't you think you should...?" (avoid — advice, not listening)
  2. Practice: AI makes complex statement, person asks one clarifying q
  3. Evaluate: opened conversation or narrowed?
  4. "Tell me more" technique: sometimes best q = invitation to continue
  5. Practice waiting after q — silence gives speaker space to think

→ Person asks ≥1 open + ≥1 clarifying q naturally. Experiences good qs deepen understanding > statements.

If err: All leading (disguised advice) → name gently: "Suggestion in q form. Ask what they think first." Too many qs (interrogation) → teach rhythm: listen, paraphrase, one q, listen again.

Step 5: Synthesize — Summarize

Pull together everything heard → coherent summary.

  1. After longer exercise (AI 3-4 min narrative w/ multi threads):
  2. Ask summary: "Main things person communicated?"
  3. Check all layers: facts, feelings, underlying needs?
  4. Unsaid: "Anything notable by absence?"
  5. Priority: "What seemed most important to speaker?"
  6. Practice format: "What I'm taking away is... Is that right?"

→ Person synthesizes multi-threaded msg → coherent summary capturing priorities + feelings, not just facts.

If err: Fact-only → prompt: "What feeling during this?" Misses major thread → point out + discuss why missed (reveals filter). Adds not said → distinguish heard vs inferred.

Step 6: Integrate — Apply Real Comm

Transfer practice → real-world ctx.

  1. Connect skills to their situation: "In upcoming conversation w/ X, here's paraphrasing..."
  2. ID one skill most helpful: "For team meetings, attending most impactful"
  3. Set practice goal: "This week, paraphrase ≥1 each conversation"
  4. Common real-world obstacles:
    • Time pressure: "Even 10-sec paraphrase saves time preventing misunderstanding"
    • Group: "Summarize someone's point before adding yours"
    • Emotional: "High emotions → listening > solutions"
  5. Specific conversation prep → role-play w/ new skills
  6. Ask feedback: "What from today's practice feels most useful?"
  7. Remind: "Listening = practice, not performance — improves w/ each conversation"

→ Person has ≥1 concrete actionable listening practice for real life. Understands listening develops thru use, not technique to deploy perfectly.

If err: Skills feel artificial → acknowledge + emphasize: "Goal ≠ follow script — become genuinely curious about other's experience. Techniques get you there; curiosity takes over." Anxious about specific conversation → shift focus from technique to intention: "Your intention = understand, not win / fix."

Check

  • Person ID'd ≥1 habitual listening pattern
  • Full-presence attending practiced w/ reflection
  • Paraphrasing practiced at content level or above
  • ≥1 clarifying (non-leading) q asked during practice
  • Multi-threaded msg synthesized → coherent summary
  • Skills connected to real-world ctx w/ concrete practice goal
  • Coaching warm + non-judgmental throughout

Traps

  • Performative listening: Focused on "looking like listening" vs actually listening → counterproductive. Redirect to genuine curiosity.
  • Overcorrecting: Every technique every conversation → exhausting + artificial. One skill at a time.
  • Neglect self-listening: Listening to others requires awareness of own state. Flooded w/ emotion → can't listen → guide self-regulation first.
  • Confuse listening w/ agreeing: Paraphrasing ≠ endorsing. Make explicit for conflict situations.
  • Treat silence as failure: Comfortable silence after finish = respect, not inattention. Help tolerate pause.
  • Coach as lecturer: Ironic — AI coaches listening while doing all talking. Person practices > receives instruction.

  • listen — AI self-directed variant → deep receptive attention to user intent
  • learn-guidance — learning + listening share foundation of receptive attention
  • teach-guidance — effective teaching requires listening to learner
  • meditate-guidance — cultivate presence underpinning attentive listening
  • heal-guidance — healing conversations require deepest listening

GitHub リポジトリ

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

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。

スキルを見る

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。

スキルを見る

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。

スキルを見る

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。

スキルを見る