clearai-dsh 0.1.0

package/presets/clearai/template/skills/what-if-oracle/SKILL.md

@@ -0,0 +1,72 @@
+---
+name: what-if-oracle
+description: |
+  【情景推演·方案比选】多分支 What-If 分析（最好/最可能/最坏/二阶效应）。适用：实验方案变更、数据源替换、方法路线切换、资源投入决策前的情景压力测试。不适用：已有数据的统计验证（用 statistical-analysis）；深度关联与根因分析（用 data-analysis）。
+license: CC BY-NC-SA 4.0
+metadata:
+  version: 1.1-clearai
+  upstream: https://github.com/ashrafkahoush-ux/claude-consciousness-skills
+  tier: system
+  origin: template
+  created_at: '2026-06-12T02:47:05.412617+00:00'
+---
+
+# What-If Oracle Skill（科研情景版）
+
+> 许可：CC BY-NC-SA 4.0，详见本目录 `LICENSE.txt`（若已复制）。
+
+## 使用边界
+
+- **适用**：决策分叉、不确定性高、需映射可能性空间而非单点预测。
+- **不适用**：可凭历史数据直接检验的假设 → `statistical-analysis`；正式预研报告 → domain/process presearch。
+
+## ClearAI 工具与路径映射
+
+- 背景事实 → `web_search` / `read` 已有报告
+- 情景文档 → `write` 到 `lab/knowledge/scenarios.md`
+- 示意图 → mermaid → `lab/diagrams/`
+- 经验回写 → `clear/memory/what_if_lessons.md`
+
+## 核心原则：精确 IF
+
+模糊「如果出问题怎么办」无效。须量化：
+- **变量**：什么变？（如 主力数据集样本量 -30%）
+- **幅度**：变多少？
+- **时间窗**：本季度 / 6 个月？
+- **当前态**：变更前基线？
+
+## 工作流
+
+### Phase 1：框定问题
+
+分解为 Variable / Magnitude / Timeframe / Context。参见 `references/scenario-templates.md` 科研/数据段落。
+
+### Phase 2：分支探索（4–6 条）
+
+| 分支 | 说明 |
+|------|------|
+| Best | 乐观但需说明条件 |
+| Likely | 基准情景 |
+| Worst | 下行风险 |
+| Wildcard | 低概率高影响 |
+| Contrarian | 与共识相反 |
+| Second-order | 一阶后果的连锁 |
+
+### Phase 3：每条分支的逻辑链
+
+- 触发条件 → 一阶影响（结果有效性/进度/成本）→ 二阶影响 → 可观测指标
+
+### Phase 4：决策建议
+
+- 不需替用户决策；列出「需监测的信号」与「建议的验证动作」（可链到 `hypothesis-generation`）
+
+## 常见场景模板
+
+- **实验方案变更**：条件/参数/流程调整对结果有效性与成本的影响
+- **数据源替换**：换用新数据集或采集渠道后的可比性与偏差风险
+- **方法路线切换**：更换分析/建模方法的收益、迁移成本与回退路径
+- **资源投入决策**：加大算力/实验投入前的收益-风险压力测试
+
+## 交付物 `lab/knowledge/scenarios.md`
+
+含分支表、概率区间（定性高/中/低即可）、监测指标、下一步验证建议。

package/presets/clearai/template/skills/what-if-oracle/references/scenario-templates.md

@@ -0,0 +1,154 @@
+# What-If Oracle — Scenario Templates
+
+Reference guide for domain-specific scenario analysis configurations.
+
+## Research / Data Science
+
+**Variables to test:**
+
+- Experiment protocol changes (conditions, parameters, sample size)
+- Data source replacement or collection interruption
+- Method/model route switch (new algorithm vs. established baseline)
+- Resource allocation (compute, personnel, funding, time window)
+- Validity threats (confounders, drift, selection bias)
+
+**Branch emphasis:** Likely Case + Worst Case + Second Order
+**Key tension:** Rigor vs. iteration speed — evidence strength vs. time-to-conclusion
+
+**Template prompt:**
+
+> "What if we [change protocol / swap data source / switch method] in [timeframe]? Current baseline: [design, sample size, key metrics]. Key constraint: [compute/funding/deadline]. What we must not compromise: [validity requirement]."
+
+## Startup / Business Decision
+
+**Variables to test:**
+
+- Market entry timing
+- Pricing strategy
+- Partnership terms
+- Hiring decisions
+- Funding rounds
+
+**Branch emphasis:** Likely Case + Contrarian + Second Order
+**Key tension:** Speed vs. thoroughness — startups can't afford to analyze forever
+
+**Template prompt:**
+
+> "What if we [specific action] in [timeframe]? Our current state: [revenue, team size, runway]. Key constraint: [the limiting factor]."
+
+## Technology Architecture
+
+**Variables to test:**
+
+- Tech stack choice
+- Build vs. buy
+- Scaling approach
+- Security model
+- Migration path
+
+**Branch emphasis:** Worst Case + Wild Card + Second Order
+**Key tension:** Engineering elegance vs. shipping speed
+
+**Template prompt:**
+
+> "What if we choose [technology/approach] for [system]? Current architecture: [brief description]. Team capability: [relevant skills]. Timeline: [deadline]."
+
+## Investment / Financial
+
+**Variables to test:**
+
+- Market conditions
+- Valuation scenarios
+- Exit timing
+- Capital allocation
+- Revenue model changes
+
+**Branch emphasis:** All 6 branches — money decisions deserve full analysis
+**Key tension:** Risk tolerance vs. opportunity cost
+
+**Template prompt:**
+
+> "What if [market condition / financial event] happens? Our exposure: [amount/percentage]. Current position: [financial state]. Time horizon: [investment period]."
+
+## Career / Personal
+
+**Variables to test:**
+
+- Job change
+- Skill development path
+- Relocation
+- Relationship decisions
+- Health changes
+
+**Branch emphasis:** Likely Case + Best Case + Contrarian
+**Key tension:** Security vs. growth — comfort zone vs. expansion
+
+**Template prompt:**
+
+> "What if I [personal decision]? My current situation: [brief]. What I value most: [1-3 values]. What I'm afraid of: [honest answer]."
+
+## Geopolitical / Macro
+
+**Variables to test:**
+
+- Policy changes
+- Regulatory shifts
+- Market disruptions
+- Technology breakthroughs
+- Social movements
+
+**Branch emphasis:** Wild Card + Second Order + Contrarian
+**Key tension:** Local impact vs. systemic effects
+
+**Template prompt:**
+
+> "What if [geopolitical event] happens? My exposure: [how it affects me/my organization]. Time horizon: [relevant period]."
+
+## Crisis Response
+
+**Variables to test:**
+
+- Severity escalation
+- Communication strategy
+- Resource allocation
+- Recovery timeline
+- Reputation impact
+
+**Branch emphasis:** Worst Case (detailed) + Likely Case + Second Order
+**Key tension:** Immediate triage vs. root cause resolution
+
+**Template prompt:**
+
+> "We're facing [crisis/incident]. Current impact: [what's broken]. Stakeholders affected: [who]. Resources available: [what we can deploy]. What if [specific escalation scenario]?"
+
+## The Recursive Template
+
+For complex, multi-layered analysis:
+
+```
+ROUND 1: "What if X?"
+  → Identify the most likely branch (α)
+
+ROUND 2: "Given α is happening, what if Y?"
+  → Identify the most likely sub-branch
+
+ROUND 3: "Given α+Y, what if Z?"
+  → Map the deepest consequences
+
+Each round narrows the possibility space while deepening understanding.
+Maximum recommended depth: 3 rounds.
+```
+
+## Probability Calibration Guide
+
+When assigning probabilities to branches:
+
+| Confidence Level | Probability Range | Evidence Required                                          |
+| ---------------- | ----------------- | ---------------------------------------------------------- |
+| **Very High**    | >80%              | Strong historical precedent + current data alignment       |
+| **High**         | 60-80%            | Multiple converging signals, some historical support       |
+| **Medium**       | 30-60%            | Mixed signals, could go either way                         |
+| **Low**          | 10-30%            | Plausible but requires several things to go a specific way |
+| **Very Low**     | <10%              | Black swan territory — possible but unlikely               |
+
+**Rule:** All branch probabilities in a single analysis should sum to approximately 100%. If they don't, there's a missing branch.