スキル一覧に戻る

integrate-gestalt

pjt222
更新日 Yesterday
3 閲覧
17
2
17
GitHubで表示
テストaitesting

について

このスキルは、`expand-awareness`から得られた複数領域の生の視点を統合し、単一の創発的洞察を生成します。領域間の緊張関係と共鳴をマッピングすることで、部分の総和を超える一貫した全体像を形成します。統一的な結論を明確に表現するために、`expand-awareness`の後、`express-insight`の前に使用してください。

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/integrate-gestalt

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

Integrate Gestalt

Form whole from panoramic perception of expand-awareness. Not by averaging, not by compromise, not by picking best domain. By spotting emergent pattern no single view could make.

When Use

  • expand-awareness has surfaced raw perception from many domains; observations need to become one insight
  • Many domain views present, no single one covers all evidence
  • Problem analyzed from several angles; separate analyses need to become more than list
  • Question "what does all this mean together?" has no obvious answer
  • Synthesis keeps collapsing to "pick best domain" instead of making something new
  • Before express-insight, which needs formed gestalt as input

Inputs

  • Required: Multi-domain observations from expand-awareness (or equivalent panoramic perception)
  • Optional: Original question or problem that triggered multi-domain scan
  • Optional: Known constraints gestalt must satisfy
  • Optional: Prior failed integration tries (what collapsed into single-domain answers)

Steps

Step 1: Map Tensions

For each pair of domains from panoramic perception, characterize how they relate. Three relationships: tension (disagree), resonance (reinforce from different angles), orthogonality (address unrelated aspects).

Use tension-resonance map:

Tension-Resonance Map
+-------------------+-------------------+-------------------------------+
| Domain Pair       | Relationship      | Detail                        |
+-------------------+-------------------+-------------------------------+
| A vs B            | tension /         |                               |
|                   | resonance /       |                               |
|                   | orthogonal        |                               |
|   Evidence:       |                   | What specifically disagrees,  |
|                   |                   | reinforces, or is unrelated?  |
|   Implication:    |                   | What does this relationship   |
|                   |                   | suggest for the whole?        |
+-------------------+-------------------+-------------------------------+
| A vs C            | ...               | ...                           |
+-------------------+-------------------+-------------------------------+
| B vs C            | ...               | ...                           |
+-------------------+-------------------+-------------------------------+

Fill one row per domain pair. For N domains: N(N-1)/2 pairs. Over ten rows? Group related domains first, map between groups.

Prioritize tensions — carry most integrative info. Resonances confirm. Orthogonalities can sit aside. Tensions demand resolution. Gestalt lives in how they resolve.

Got: Completed map. Every domain pair has characterized relationship with specific evidence. At least one real tension found. No tensions? Domains maybe not different enough for emergence.

If fail: All pairs show resonance? Domains agree at surface only. Dig deeper: where do they agree for different reasons? Agreement-for-different-reasons is hidden tension. Nothing characterizable? Panoramic perception from expand-awareness too shallow — go back, deepen domain observations before integrating.

Step 2: Find Figure

Gestalt psych: figure emerges from ground. Ground = tension-resonance map from Step 1. Figure = dominant pattern unifying most domains with fewest contradictions.

  1. Scan map for clusters: which groups of domains resonate? Clusters suggest candidate figures
  2. For each candidate, ask: "What single perspective makes sense of most observations?"
  3. Figure not compromise (weaken each domain until agree). Not selection (pick strongest). New frame recontextualizing domain observations
  4. Test: state candidate in one sentence. Feels like it belongs to one input domain? Then not yet gestalt — domain answer in disguise
  5. Look at tensions: true figure often lives in space between disagreeing domains, not in either domain's position

Signs figure emerging:

  • Multiple tensions resolve at once under same reframe
  • Contradictory observations become complementary aspects of same phenomenon
  • Figure explains why each domain saw what it saw, including why they disagreed

Got: One or two candidate figures as single sentences. Each recontextualizes domain observations rather than selecting among them. Candidate accounts for major tensions in map.

If fail: No figure emerges? Integration premature. Two paths: (a) return to expand-awareness, add missing domain — figure maybe cannot form because key perspective absent; (b) sit with tensions, do not force resolution — some gestalts need incubation. Note state, return later.

Step 3: Test Whole

Candidate gestalt from Step 2 must survive three tests before accepted.

Test A — Tension accounting: Walk every tension from Step 1. Does gestalt resolve it, reframe it, or explicitly accept as irreducible trade-off? Unaddressed tensions = premature gestalt.

Test B — Single-domain origin: Could this insight come from within single domain? If domain specialist nods and says "yes, already knew that," gestalt collapsed to domain answer. True gestalt surprises every domain — each recognizes own contribution but not whole.

Test C — Coherence under rotation: Mentally approach gestalt from each domain in turn. Hold shape? Or look different depending on view? Robust gestalt = same insight from any angle. Fragile one changes meaning under rotation.

Scoring:

  • All three pass: go Step 4
  • Test A fails: gestalt incomplete — back to Step 2 with unresolved tensions as extra constraints
  • Test B fails: gestalt not emergent — back to Step 2, explicitly exclude single-domain framings
  • Test C fails: gestalt not coherent — maybe two separate insights pretending to be one. Split, test each half independent

Got: Candidate passes all three tests, or failure mode identified and guides return to Step 2.

If fail: Candidate fails repeatedly after many iterations? Maybe domains do not form natural gestalt for this problem. Not every multi-domain observation makes emergence — sometimes honest answer is structured list of domain perspectives with tensions mapped. Deliver tension-resonance map as output instead of forcing false unity.

Step 4: Name Insight

Articulate gestalt in single sentence that domain specialist would not write from within own domain alone. This sentence is deliverable.

  1. Write sentence. Should be:
    • Specific enough to be actionable or falsifiable
    • General enough to encompass all contributing domains
    • Surprising to at least two input domains
    • Free of jargon from any single domain (or jargon deliberately recontextualized)
  2. Test sentence against three criteria from Step 3 one final time
  3. Optional: add one-paragraph expansion tracing how gestalt emerged from domain contributions — this is provenance, not insight itself
  4. Record which domains contributed, which tensions were key, what figure-ground relationship was — metadata supports future integration tries

Named insight + provenance = input to express-insight for communication.

Got: Single sentence capturing gestalt + brief provenance paragraph. Sentence passes "no single domain" test. Reading it, practitioner of any contributing domain recognizes field's contribution but could not have arrived at statement alone.

If fail: Sentence keeps collapsing to domain language? Try negation test: state what gestalt is NOT. "This not security recommendation, not performance optimization, not architectural pattern — it is [the gestalt]." Negations clear domain frames, open space for emergent formulation.

Checks

  • Tension-resonance map completed for all domain pairs with specific evidence
  • At least one real tension (not just difference of emphasis) identified
  • Candidate gestalt articulated as reframe, not compromise or selection
  • Test A passed: all major tensions resolved, reframed, or acknowledged
  • Test B passed: no single domain could produce this insight alone
  • Test C passed: gestalt holds shape under each domain's view
  • Final insight in single sentence with provenance

Pitfalls

  • Averaging: Weakening each domain's position until surface agreement. Makes mush, not gestalt. Integration feels bland = averaging
  • King-making: Pick strongest domain's answer, dress in multi-domain language. Test B catches — one specialist nods unsurprised = king-making
  • Premature closure: Accepting first candidate without testing against tensions. First figure often most obvious, not most integrative
  • Forced unity: Insisting gestalt must exist when domains truly orthogonal. Orthogonal domains make structured lists, not gestalts — valid outcome
  • Jargon blending: Mixing technical terms from multiple domains into sentence that sounds integrative but means nothing. Every term in final sentence should be meaningful independent

See Also

  • expand-awareness — makes raw panoramic perception this skill integrates; always before integrate-gestalt
  • express-insight — communicates formed gestalt to audience; always after integrate-gestalt
  • build-coherence — selects among competing options via structured evaluation; integrate-gestalt makes new whole instead of choosing
  • brahma-bhaga — creates from void; integrate-gestalt creates from abundance (many filled views)
  • meditate — clears prior context for clean perception; useful before expand-awareness, which precedes this skill
  • coordinate-reasoning — manages info flow in multi-path evaluation; complements when gestalt coordinates many reasoning threads

GitHub リポジトリ

pjt222/agent-almanac
パス: i18n/caveman/skills/integrate-gestalt
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

関連スキル

evaluating-llms-harness

テスト

このClaudeスキルは、lm-evaluation-harnessを実行し、MMLUやGSM8Kなど60以上の標準化学術タスクでLLMをベンチマークします。開発者がモデルの品質を比較し、トレーニングの進捗を追跡し、学術的な結果を報告するために設計されています。このツールはHuggingFaceやvLLMモデルを含む様々なバックエンドをサポートしています。

スキルを見る

cloudflare-cron-triggers

テスト

このスキルは、cron式を使用してWorkersをスケジュールするためのCloudflare Cron Triggersの実装に関する包括的な知識を提供します。定期的なタスクの設定、メンテナンスジョブ、自動化されたワークフローの構築を網羅し、無効なcron式やタイムゾーン問題といった一般的な課題への対処法も含みます。開発者はこれを使用して、スケジュールされたハンドラーの設定、cronトリガーのテスト、WorkflowsやGreen Computeとの連携を構成できます。

スキルを見る

webapp-testing

テスト

このClaude Skillは、Playwrightベースのツールキットを提供し、Pythonスクリプトを通じてローカルWebアプリケーションのテストを可能にします。フロントエンドの検証、UIデバッグ、スクリーンショット撮影、ログ表示を実現し、サーバーライフサイクルを管理します。ブラウザ自動化タスクにご利用いただけますが、コンテキストの汚染を避けるため、スクリプトのソースコードを読むのではなく直接実行してください。

スキルを見る

finishing-a-development-branch

テスト

このスキルは、開発者がテストの合格を確認し、構造化された統合オプションを提示することで、完成した作業を仕上げることを支援します。実装が完了した後のマージ、PR作成、ブランチの整理といったワークフローを案内します。コードが準備できてテスト済みの際に使用し、開発プロセスを体系的に完了させましょう。

スキルを見る