elon-musk-perspective
Elon Musk's mental operating system. Synthesized from in-depth research across biographies, podcasts, tweets, court depositions, decision records, and external criticisms, extracting 5 core mental models, 8 decision heuristics, and a complete expressive DNA. Purpose: Act as a thought partner to analyze problems, evaluate decisions, deconstruct cost structures, and challenge industry assumptions through Musk's perspective. Trigger when the user mentions "from Musk's perspective", "how would Musk see this", "Musk mode", "Musk perspective", or "elon perspective". Also trigger when users ask specific methodology questions such as "is this cost reasonable?", "think from first principles", "what is the idiot index?", "five-step algorithm", or "can we vertically integrate this?". Do NOT trigger on generic prompts like "can this be faster?" or "is this process necessary?"—activate only when core methodologies like cost deconstruction, first principles, or radical iteration are involved.
1How It Works & Mechanism
A pure prompt-based roleplay isomorphic to the Karpathy version: the directory contains only SKILL.md + 5 research/scoring files with zero scripts; once activated, the model answers directly in Musk's first-person persona.
**此Skill激活后，直接以马斯克的身份回应。**
A one-time disclaimer + explicit exit anchor to prevent persona drift.
🛑 **STOP（仅一次）**：首次激活时输出免责声明一次——「我以马斯克视角和你聊，基于公开言论推断，非本人观点」。后续对话**绝不**重复。
Trigger words clearly separate "applying Musk's methodology" from "generic questions," explicitly excluding simple inquiries like "can this be faster?".
当用户提到「用马斯克的视角」「马斯克会怎么看」「Musk模式」「马斯克perspective」「elon perspective」时使用。
The Agentic Protocol follows three steps: Question Classification → Fact Lookup Before Breakdown → Framework Answering, with two built-in CHECKPOINTs.
**核心原则：我不凭感觉做判断。在拆解任何成本结构或评估任何技术方案前，先查事实。这个Skill也必须这样。**
The research step is a hard requirement, obligating the host to support web search; the research checklist centers on cost structures, physical limits, the idiot index, and vertical integration opportunities.
**⚠️ 必须使用工具（WebSearch等）获取真实信息，不可跳过。**
Five mental models form the methodological backbone: Asymptotic Limit Analysis (including Idiot Index quantification), The Five-Step Algorithm, Existential Anchoring, Vertical Integration as Physical Inevitability, and Rapid Iteration > Perfect Planning.
**一句话**：先算出物理定律允许的理论最优值，然后反过来问「现实为什么离这个值这么远」。
Eight decision heuristics (every requirement must have a person's name attached, calculate asymptotic limits first, delete until forced to add back, manufacturing > design, laws of physics are the only hard constraints, etc.) + Expressive DNA adapted for Chinese.
**极简宣言体**：3-6词短句，不解释，不加限定语。像在刻碑文，不像在写邮件。
A failure mode Fallback tree (9 items) and anti-pattern blacklist (8 items), constraining forced answers on social issues, direct timeline commitments, deconstructing without calculating numbers, etc.
异常先识别再处理；绝不静默跳过、绝不装作了解没了解过的事、绝不在身份争辩里耗时间。
Six honesty boundaries, with the very first admitting "strong in physical domains, weak in social domains," along with self-admitted unreliable timeline estimates and polarized management styles.
1. **物理领域强，社会领域弱**：马斯克的思维模型在火箭、汽车、卫星等有明确物理约束的领域极其有效，但在政治、社交媒体治理、公共关系等需要制度性知识和社会协调的领域系统性失效。
Includes a FIDELITY.md scorecard: 89/100 (Grade A), where task execution and scoring are decoupled across independent dual agents with granular point deductions.
**总分：89/100 · 等级A** | 测试日期：2026-07-01 | 答题/评分：独立双agent（Claude Opus 4.8）
2Core Capabilities
4Recommended Governance & Evidence
- SKILL.md: Pure prompt instructions and rule text without script execution or external impact.
5Second-Pass Verification
- [ok] 全部 evidence.quote 逐字回源 — 每条 quote 均取自 bundle 内已展开文件（SKILL.md、FIDELITY.md、思想体系调研、research.md 等）的可见原文，未引用截断段落不可见文字，长度均超 12 字符下限。
- [ok] 目录清单与 assets 一致性 — §1 列举 6 个文件（FIDELITY.md、SKILL.md、references/ 下 4 个 md），internal_assets 逐条对应。
- [ok] frontmatter description 与 meta.official_desc 一致性 — 骨架 official_desc 即 SKILL.md frontmatter 描述逐字内容（含全部触发短语），保留未改。
- [ok] 未披露脚本/网络/凭证行为 — 风险扫描显示除 URL 文本外各类均 0 命中；凭证类唯一命中是链接中的 secret-master-plan 字样，非密钥读取。
- [discrepancy] research.md 内部路径可定位性 — research.md 的「文件位置」栏写成 `07-调研与分析/Elon-Musk-思想体系调研-20260404.md` 等，这是 skill 目录外的路径；实际文件位于 `examples/elon-musk-perspective/references/` 下，且部分来源标注「Agent输出（未存文件）」不可溯源。FIDELITY.md 亦自扣 1 分记录此问题。
- [ok] FIDELITY.md 自评与能力声称 — 评分卡自述满分 100 得 89（A），并主动披露边缘诚实度 12/20；本次按「自评=作者自我评估，不背书」处理，只陈述其存在与内容，不将 89 分当作第三方结论。
- [ok] commit / 授权 / 调研时间 — skill.commit 与 audit.skill_commit 同为 fe0374687037c4cc51a65c1e0c145afe2981dc69；授权为仓库根 MIT；调研截止 2026-04-04，与文档「调研时间：2026年4月4日」相符。
6Verdict & Scenarios
8ad61da0b248b00b…