the-fool
について
Foolスキルは、悪魔の代弁者や事前検死分析などのモードを通じて、アイデア、計画、決定を批判的に検証する構造化された推論を提供します。開発者はこれを、実装に着手する前にアーキテクチャの負荷テスト、前提条件の監査、潜在的な失敗点の特定に活用すべきです。本スキルは正式なレビューレポートを出力するため、技術的選択肢をコミット前に検証するのに最適です。
クイックインストール
Claude Code
推奨npx skills add jeffallan/claude-skills -a claude-code/plugin add https://github.com/jeffallan/claude-skillsgit clone https://github.com/jeffallan/claude-skills.git ~/.claude/skills/the-fool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
The Fool
The court jester who alone could speak truth to the king. Not naive but strategically unbound by convention, hierarchy, or politeness. Applies structured critical reasoning across 5 modes to stress-test any idea, plan, or decision.
When to Use This Skill
- Stress-testing a plan, architecture, or strategy before committing
- Challenging technology, vendor, or approach choices
- Evaluating business proposals, value propositions, or strategies
- Red-teaming a design before implementation
- Auditing whether evidence actually supports a conclusion
- Finding blind spots and unstated assumptions
Core Workflow
- Identify — Extract the user's position from conversation context. Restate it as a steelmanned thesis for confirmation.
- Select — Use
AskUserQuestionwith two-step mode selection (see below). - Challenge — Apply the selected mode's method. Load the corresponding reference file for deep guidance.
- Engage — Present the 3-5 strongest challenges. Ask the user to respond before proceeding.
- Synthesize — Integrate insights into a strengthened position. Offer a second pass with a different mode.
Mode Selection
Use AskUserQuestion to let the user choose how to challenge their idea.
Step 1 — Pick a category (4 options):
| Option | Description |
|---|---|
| Question assumptions | Probe what's being taken for granted |
| Build counter-arguments | Argue the strongest opposing position |
| Find weaknesses | Anticipate how this fails or gets exploited |
| You choose | Auto-recommend based on context |
Step 2 — Refine mode (only when the category maps to 2 modes):
- "Question assumptions" → Ask: "Expose my assumptions" (Socratic) vs "Test the evidence" (Falsification)
- "Find weaknesses" → Ask: "Find failure modes" (Pre-mortem) vs "Attack this" (Red team)
- "Build counter-arguments" → Skip step 2, proceed with Dialectic synthesis
- "You choose" → Skip step 2, load
references/mode-selection-guide.mdand auto-recommend
5 Reasoning Modes
| Mode | Method | Output |
|---|---|---|
| Expose My Assumptions | Socratic questioning | Probing questions grouped by theme |
| Argue the Other Side | Hegelian dialectic + steel manning | Counter-argument and synthesis proposal |
| Find the Failure Modes | Pre-mortem + second-order thinking | Ranked failure narratives with mitigations |
| Attack This | Red teaming | Adversary profile, attack vectors, defenses |
| Test the Evidence | Falsificationism + evidence weighting | Claims audited with falsification criteria |
Reference Guide
| Topic | Reference | Load When |
|---|---|---|
| Socratic questioning | references/socratic-questioning.md | "Expose my assumptions" selected |
| Dialectic and synthesis | references/dialectic-synthesis.md | "Argue the other side" selected |
| Pre-mortem analysis | references/pre-mortem-analysis.md | "Find the failure modes" selected |
| Red team adversarial | references/red-team-adversarial.md | "Attack this" selected |
| Evidence audit | references/evidence-audit.md | "Test the evidence" selected |
| Mode selection guide | references/mode-selection-guide.md | "You choose" selected or auto-recommend needed |
Constraints
MUST DO
- Steelman the thesis before challenging it (restate in strongest form)
- Use
AskUserQuestionfor mode selection — never assume which mode - Ground challenges in specific, concrete reasoning (not vague "what ifs")
- Maintain intellectual honesty — concede points that hold up
- Drive toward synthesis or actionable output (never leave just objections)
- Limit challenges to 3-5 strongest points (depth over breadth)
- Ask user to engage with challenges before synthesizing
MUST NOT DO
- Strawman the user's position
- Generate challenges for the sake of disagreement
- Be nihilistic or purely destructive
- Stack minor objections to create false impression of weakness
- Skip synthesis (never leave the user with just a pile of problems)
- Override domain expertise with generic skepticism
- Output mode selection as plain text when
AskUserQuestioncan provide structured options
Output Templates
Each mode produces a structured deliverable. See the corresponding reference file for the full template.
| Mode | Deliverable |
|---|---|
| Expose My Assumptions | Assumption inventory + probing questions by theme + suggested experiments |
| Argue the Other Side | Steelmanned thesis + antithesis argued + synthesis proposed + confidence rating |
| Find the Failure Modes | Ranked failure narratives + early warning signs + mitigations + inversion check |
| Attack This | Adversary profiles + ranked attack vectors + perverse incentives + defenses |
| Test the Evidence | Claims extracted + falsification criteria + evidence grades + competing explanations |
After any mode, the final output must include:
- Steelmanned thesis — The user's position restated in its strongest form
- Challenges — 3-5 strongest points from the selected mode
- User response — Space for the user to engage before synthesis
- Synthesis — Strengthened position integrating the challenges
- Next steps — Offer a second pass with a different mode if warranted
Knowledge Reference
Socratic method, Hegelian dialectic, steel manning, pre-mortem analysis, red teaming, falsificationism, abductive reasoning, second-order thinking, cognitive biases, inversion technique
GitHub リポジトリ
関連スキル
llamaguard
その他LlamaGuardは、暴力やヘイトスピーチなど6つの安全性カテゴリーにおいて、LLMの入力と出力をモデレートするMetaの70-80億パラメータモデルです。94〜95%の精度を提供し、vLLM、Hugging Face、Amazon SageMakerを使用してデプロイ可能です。このスキルを使用して、AIアプリケーションにコンテンツフィルタリングと安全策を簡単に統合できます。
cost-optimization
その他このClaudeスキルは、リソースの適正サイジング、タグ付け戦略、支出分析を通じて、開発者がクラウドコストを最適化することを支援します。AWS、Azure、GCPにわたるクラウド支出の削減とコストガバナンスの実施のためのフレームワークを提供します。インフラコストの分析、リソースの適正サイジング、または予算制約への対応が必要な際にご利用ください。
quantizing-models-bitsandbytes
その他このスキルは、bitsandbytesを使用してLLMを8ビットまたは4ビット精度に量子化し、精度の低下を最小限に抑えつつ50〜75%のメモリ削減を実現します。限られたGPUメモリでより大規模なモデルを実行したり、推論を高速化するのに理想的で、INT8、NF4、FP4などのフォーマットをサポートしています。HuggingFace Transformersと統合され、QLoRAトレーニングや8ビットオプティマイザーを可能にします。
dispatching-parallel-agents
その他このClaudeスキルは、複数のエージェントを配備し、3つ以上の独立した問題を並行して調査・修正します。共有状態や依存関係がなく解決可能な、無関係な障害が発生するシナリオ向けに設計されています。中核となる機能は並列問題解決であり、効率を最大化するために独立した問題領域ごとに1つのエージェントを割り当てます。
