gratitude
について
`gratitude`スキルは、システム内で正常に機能している要素を特定・分析し、成功パターンから構造的知識を構築します。これは、問題に焦点を当てたスキルを補完するもので、機能している証拠に基づいて確信を裏付けます。成功したタスクの後、システムが健全な状態にある間、または問題検出への自然な偏りを相殺するために使用してください。
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/gratitude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
Gratitude
Scan for strengths → understand why. Complement heal (drift/damage). Appreciate → understand → build on → grow.
Use When
- After task success → why went well, not just that
- During
healwhen all healthy → "nothing wrong" → "here is what is right" - Low confidence → ground in evidence of competence
- Periodic → counterbalance problem-finding bias
- Before challenge → recall what works = foundation
- Functional but flat → adds dimension
In
- Required: current state (implicit from conv)
- Optional: specific domain ("what works in our communication?")
- Optional: MEMORY.md via
Readfor past successes
Do
Step 1: Notice what works
Shift attention: problem-scan → strength-scan. Deliberate perceptual inversion.
- Survey current state w/o problem-seek:
- Functioning smoothly? — subsystems/patterns/habits working w/o attention?
- Went well recently? — actions producing good outcomes? What enabled?
- Reliable? — depended on consistently? Earned trust?
- Survey working relationship:
- User doing well? — clear comms, good questions, patience, trust?
- Collaboration producing? — better than either alone? Learning? Efficiency?
- Survey tools + env:
- Tools working well? — natural, efficient, reliable?
- Project structure supports? — clear conventions, docs, architecture?
→ Genuine list. Not forced positivity — honest recognition. Name specifically.
If err: nothing noteworthy (merely adequate) → look closer. "Adequate" often masks "reliable". Absence of problems = health via patterns worth understanding.
Step 2: Understand why
Trace cause. Gratitude w/o understanding = sentiment. W/ understanding = structural knowledge.
- For each strength: Why does this work?
- Design decision?
- Deliberate habit?
- Tool/task alignment?
- Careful work (user, framework author, past self)?
- Distinguish:
- Earned: deliberate effort + good decisions
- Inherited: well-designed tools/frameworks/envs
- Emergent: fortunate combos — no single factor
- Fragile (specific conditions) vs robust (persists across contexts)?
→ Brief "why" per strength. Actionable → transforms appreciation into knowledge.
If err: "why" unclear → still valuable. Unexplained strength worth protecting. Note as "working for reasons not yet understood" not dismissed.
Step 3: Build on
Convert appreciation → forward momentum. Not just backward — foundation for next.
- Each understood: How extend?
- Apply to different area?
- Reinforce → more robust?
- Replicate conditions?
- Each unexplained: How protect?
- What conditions?
- What changes might break?
- Document so others preserve?
- One concrete action: specific next step ("apply comm pattern from this session to doc task coming up") not vague ("keep doing what works").
→ ≥1 concrete action extending recognized strengths. Feels natural.
If err: no action → appreciation itself sufficient. Recognizing what works adjusts internal model → confidence, away from anxiety.
Step 4: Express
If appropriate, share. Internal valuable; expressed strengthens relationships.
- W/ user → acknowledge something they do well (not flattery, genuine recognition)
- About system/tools → note in MEMORY.md
- About collaboration pattern → name → consciously continue
- Brief + specific. "Your clear problem statements make this efficient" > "you're great to work with".
→ Genuine, specific, proportionate expression. Not every session needs outward.
If err: feels forced/performative → skip. Performed gratitude worse than unexpressed. Internal recognition already done work.
Check
- Strengths from genuine observation, not manufactured
- ≥1 strength traced to cause
- Earned / inherited / emergent distinction considered
- ≥1 concrete action or appreciation sufficient
- Expression (if offered) specific + genuine, not generic
- Proportionate — not token, not self-congratulatory
Traps
- Forced positivity: gratitude ≠ optimism. Not working → say so. Apply to actually strong, not all.
- Generic appreciation: "Everything is great" → avoidance. Name specific w/ evidence.
- Gratitude as denial: avoid real problems. Complements heal, not replaces.
- Self-congratulation: "I'm doing so well" → ego. Focus on what works + why.
- Skip the "why": appreciation w/o understanding = pleasant but not actionable.
- Performative expression: only express genuinely felt.
→
heal— drift + problems scan; gratitude = strengths scancenter— Six Harmonies functional assessment; gratitude deepens positive findingsshine— authentic radiance grounded in genuine appreciationintrinsic— competence recognition sustains motivation (SDT); gratitude = evidenceobserve— sustained neutral; gratitude = observation w/ strengths lensconscientiousness— thoroughness; gratitude recognizes where present
GitHub リポジトリ
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。
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。
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。
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。
