📊 Live Dashboard · EN · 中文
119 files. 21 distilled analysts. 12 decision-maker behavioral models. 7 Persian empire mental models spanning 2,500 years. 8 mandatory bias scans. 1 devil's advocate protocol. 1 multi-agent adversarial debate system.
Load it into Claude. Ask any Iran question. Get an analysis with deeper cultural context than any think tank report and sharper probability estimates than Polymarket.
We found Polymarket overprices the US-Iran nuclear deal by 10-15 percentage points — because the market assumes Iran will "rationally" accept economic incentives. It won't. When the US says "give up enrichment," Iran hears "surrender your sovereignty like your ancestors did under the Qajar dynasty in 1901." That's not strategy — it's 125 years of trauma, coded into political DNA.
| Dimension | Market wins | Skill wins |
|---|---|---|
| 48-hour predictions | ✅ | |
| "What will Iran accept?" | ✅ Cultural depth = alpha | |
| Netanyahu cross-market causality | ✅ Spoiler mechanism modeled | |
| Frozen conflict pricing | ✅ Markets can't price gray zones | |
| Black swans | ✅ |
The alpha isn't speed — it's depth. Every time the market's implicit assumption is "rational Iran should accept this deal," 2,500 years of sovereignty trauma is the alpha source.
Iran.skill/ (119 files, 390KB)
├── SKILL.md ← Controller (530 lines)
├── /sources/ ← 32 analyst knowledge bases (Sadjadpour, Takeyh, Nasr, Vaez...)
├── /decisionmakers/ ← 12 behavioral models (Trump, Mojtaba, Netanyahu, Xi...)
├── /perspectives/ ← 30 activatable thinking lenses
└── /history/ ← 27 files — the unique weapon
├── /leaders/ ← 7 empire mental models (Cyrus → Khomeini)
├── /frameworks/ ← 8 analytical frameworks
│ Israel spoiler mechanism · Economic transmission chain
│ Society pulse · Narrative engine (Wuxia / Shakespeare)
├── /us-patterns/ ← US war decision patterns (Trump = Jackson + Nixon)
└── /ops/ ← Red Team · Adversarial Debate · Probability audit
Forecasting toolkit · Mojtaba tracking · 10 Black Swans
When a single-perspective analysis isn't enough, the skill escalates to structured multi-agent debate — 3-5 distilled analysts argue over 5 rounds with enforced rules:
Round 1: Free statements → initial positions + probabilities
Round 2: Direct clash → must quote and rebut a specific opponent argument
Round 3: Evidence duel → concrete data, precedents, real-time signals
Round 4: Steel Man → restate opponent's STRONGEST case, then explain why you still disagree
Round 5: Synthesis → convergence points, unresolved disputes, conditional judgments
Not theater — a convergence mechanism. Each round forces participants to engage with opposing evidence. The most valuable output isn't the final probability — it's the disagreement map: "If X happens, Sadjadpour is right. If Y happens, Parsi is right."
Try it: "Let Sadjadpour and Parsi debate whether Iran will accept a nuclear deal by June."
- ⚖️ Source bias labeled — funding and ideology disclosed for every source
- ⏱ Temporal partition — pre-war vs post-war analysis separated
- 🔬 Prediction type declared — structural (good at) vs engineering (bad at, must disclose)
- 🔲 4-angle verification — US / Iran / third-party / historical pattern
- 🌐 Language blind spot flagged — English sources lag Persian/Arabic by 2-6 hours
- ⛔ 6 proven-wrong assumptions auto-blocked — "protests = regime collapse" etc.
- 🔄 Decision-maker models expire in 72 hours during wartime
- 📖 Historical analogies cannot be over-simplified — 2026 Iran ≠ Qajar dynasty
| Scenario | Probability | Confidence |
|---|---|---|
| Korea-style frozen conflict (180d) | 50-60% ★ | MC |
| Framework deal (90d) | 35-45% | MC |
| Ceasefire collapse (30d) | 15-20% | LC |
| Regime change | <8% | HC (falsified) |
Simulated review from Brookings Institution perspective: B+/A-
"Its greatest value is not giving the right answer — but ensuring that before you give an answer, you've asked all the questions that should be asked."
5 expert reviews with critiques and our responses → EXPERT-FEEDBACK.md
1. Clone this repo
2. Create a Claude Project → upload all files
3. Ask any Iran question
4. Claude auto-executes: real-time search → historical matching → 4-angle verification → bias scan → Red Team → outputThe methodology is replicable. Swap Iran for any crisis:
- Ukraine.skill — Soviet collapse mental models + NATO expansion + Russian strategic culture
- Taiwan.skill — Chinese imperial cycle + US-China precedents + Japan/Korea variables
Core components (directly reusable): 女娲 distillation · 4-angle verification · Historical pattern matching · Red Team protocol · Multi-agent adversarial debate · Probability audit trail · Narrative engine
Roadmap: If Iran.skill proves its value, the architecture will generalize into a Global Geopolitical Analysis OS — same methodology, multiple crisis modules, cross-crisis interaction modeling. Iran.skill is the proof of concept. Star this repo to follow the evolution.
「江湖上有一条海峡，二十一里宽。四十七天前，它被封死了。」
119个文件。21位蒸馏分析师。7个波斯帝国心智模型。能告诉你为什么伊朗拒绝「零浓缩」——因为恺加王朝在1901年交出的石油权至今没被还回来。能告诉你Trump下一步会做什么——因为Andrew Jackson在1830年代就做过同样的事。能告诉你Netanyahu为什么在停火后几小时轰炸黎巴嫩——因为战争结束=选举=入狱。
把它喂给 Claude，问任何伊朗问题。你会得到一个比智库更有文化纵深、比 Polymarket 更准确的分析。
| 决策者 | 武侠 | 核心逻辑 |
|---|---|---|
| Trump | 曹操 | 百万大军过不了江——降龙十八掌打不死凌波微步 |
| Mojtaba | 慕容复 | 继承了一个「复国」使命但资源不够，被IRGC绑架 |
| Netanyahu | 岳不群 | 借刀杀人——打黎巴嫩→伊朗关海峡→Trump怪伊朗 |
| Xi | 司马懿 | 吸星大法——给伊朗刚好够活但不够赢，等所有人打累了 |
| IRGC | 诸葛亮 | 凌波微步——打不过你但锁住你的粮道 |
| 伊朗 | 乔峰 | 困局：无论做什么，身份标签不变。放弃浓缩也是威胁，不放弃也是威胁。 |
当单一视角不够时，skill 自动升级为结构化对抗性辩论——让蒸馏的分析师真正交锋：
第1轮：自由陈述 → 各人初始概率
第2轮：直接交锋 → 必须引用对方原话反驳，不能自说自话
第3轮：证据对决 → 拿数据/先例/实时信号说话
第4轮：钢人论证 → 用最强方式重述对手的论点，然后解释为什么自己仍然对
第5轮：综合 → 收敛点 + 分歧图 + 条件判断
最重要的输出不是概率——是分歧图：「如果以黎谈判失败 → Mead 是对的；如果第二轮美伊谈判启动 → Parsi 是对的。」
试试：「让 Sadjadpour 和 Parsi 辩论停火前景」
1901达西石油 → 1907英俄瓜分 → 1953 CIA政变 → 1979革命
→ 1988击落客机 → 2015 JCPOA → 2018撕毁 → 【2026空袭+暗杀领袖】
每次新创伤激活全部前次记忆。「零浓缩」= 新版达西石油特许权。这就是为什么伊斯兰堡谈判桌上美方提20年、伊朗只接受5年——不是数字问题，是灵魂问题。
| 时代 | 领导人 | 模型 | 当代映射 |
|---|---|---|---|
| BC 559 | 居鲁士 | 包容性帝国 | 多族群管理传统 |
| AD 1588 | 阿巴斯一世 | 教派国家建构 | 伊斯兰共和国=萨法维2.0 · IRGC=现代奴隶兵 |
| AD 1789 | 恺加王朝 | 屈辱记忆 | 「零浓缩」过敏的根源 |
| AD 1951 | 摩萨台 ★★★ | 民主中断 | 1953=所有不信任的源代码 |
| AD 1979 | 霍梅尼 ★★★ | 革命意志 | 「喝毒酒」退出机制 → A路径的文化基础 |
Polymarket 核协议定价 54.5%，我们估计 35-45%。差额来源：市场低估了浓缩权利在伊朗政治中的身份认同含义。
| 场景 | 概率 | 武侠式一句话 |
|---|---|---|
| 冻结冲突 ★ | 50-60% | 天龙八部的结局——每个人都求不得 |
| 框架协议 | 35-45% | 双方同时宣布自己赢了，只要没人追问细节 |
| 停火崩溃 | 15-20% | 岳不群再次得手 |
从 Ferdowsi 的《列王纪》（一千年前的诗）到 Nasr 的《Iran's Grand Strategy》（2024），从20卷波斯语《伊朗通史》到 Satrapi 的《Persepolis》图像小说。5层分级，含波斯语来源。不是书单推荐——是 Skill 的心智模型提取源。
Sadjadpour(Carnegie) + Vaez(ICG) + Mead(Hudson) + Nasr(SAIS) + CSET(Georgetown) → 评分 B+/A- → 完整反馈
Iran.skill 是第一个模块。如果验证有效，架构将升级为全球地缘政治分析OS——同一套方法论，多个危机模块，跨危机交互建模。Star 这个仓库来跟踪进展。
1. 克隆本仓库
2. 创建 Claude Project → 上传所有文件
3. 问任何伊朗问题（或说「用武侠写一篇霍尔木兹分析」）
4. Claude 自动执行完整分析管线
Most needed: Persian/Arabic sources · Israeli domestic politics · Quantitative economic data · Translations → CONTRIBUTING.md
Methodology / 方法论: 女娲 · Skill by @alchaincyf
Analysis informed by / 分析框架源自: Sadjadpour · Takeyh · Nasr · Vaez · Mousavian · Parsi · Maloney · Albright · Mead · Lebovic · Janis — and Ferdowsi, a poet from a thousand years ago whose Shahnameh still shapes how Iranians understand the world.
分析框架源自30+位学者，以及Ferdowsi——一千年前的诗人，他写的《列王纪》至今影响着伊朗人理解世界的方式。
"Its greatest value is not giving the right answer — but ensuring that before you give an answer, you've asked all the questions that should be asked."
「没有止损线的预测不是预测——是信仰。没有人物的分析不是分析——是表格。」