Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →AI Berkshire skill: 行业投资研究：产业链全景扫描 + 四大师个股分析框架. Source: skills/industry-research.md.
.claude/skills/xbtlin-industry-research/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 104% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -48% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -42% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -27% | 0% |
| case-10 | ✓→✗ | ▼ Worse | -6% | 0% |
This skill is generated from skills/industry-research.md so Claude Code and Codex users share one canonical workflow.
$ARGUMENTS as the user's request in the current Codex thread.tools/ in this repository. Prefer running commands from the repository root with paths like python3 tools/financial_rigor.py ...; if the current thread starts outside the repo, locate the actual checkout path first instead of assuming a fixed home-directory path.date command to confirm today's date; treat it as the baseline for "latest" data and state the data cutoff date in the report header. Never assume the current date from training data.AGENTS.md: cross-check financial data, use exact arithmetic tools for valuation/math, and clearly label uncertainty and source gaps.对 $ARGUMENTS 行业进行系统化产业链投资研究。
从一个投资主题/逻辑链出发，完成：
用箭头链路表达从"底层趋势"到"受益标的"的因果关系，例如：
底层趋势 A
→ 导致需求 B
→ 创造瓶颈/刚需 C
→ 受益产业链 D对逻辑链的每个箭头提出质疑并寻找证据：
| 环节 | 核心假设 | 验证方式 | 数据来源 | |------|---------|---------|---------| | A→B | | 搜索行业数据/预测 | | | B→C | | 搜索供需分析 | | | C→D | | 搜索实际案例/签约 | |
列出支撑该逻辑链的已签约/已落地的真实商业事件（而非预测），例如大公司的采购协议、政策文件、行业报告等。
将行业拆解为上游→中游→下游→辅助环节，例如：
上游：原材料/资源开采 → 材料加工/提纯
中游：核心设备制造 → 系统集成/工程建设 → 新技术研发
下游：运营/服务 → 终端客户
辅助：检测/认证 → 维护服务 → 金融工具（ETF/信托）对每个环节标注：
| 环节 | 商业模式 | 毛利率区间 | 竞争格局 | 壁垒类型 | 周期性 | |------|---------|-----------|---------|---------|--------| | | 卖资源/卖设备/卖服务/收租 | | 垄断/寡头/充分竞争 | 资源/牌照/技术/规模 | 强/中/弱 |
识别产业链中供给最紧张、替代最难、利润率最高的环节——这些往往是最佳投资标的所在。
行业研究中，AI数据偏见会以独特方式放大：
行业级偏见： | 偏见类型 | 表现 | 应对 | |---------|------|------| | 成熟行业偏好 | 传统行业（银行/能源/消费）资料极多，AI分析看起来"更确定" | 确定性来自商业模式，不来自研报数量 | | 新兴行业低估 | 新行业（AI应用/合成生物等）资料少，AI分析偏保守 | 用"终局思维"而非"当前数据"判断行业价值 | | 龙头偏好 | 大公司资料远多于小公司，AI天然倾向推荐龙头 | 小公司可能有更好的风险回报比，不要因为AI分析篇幅短就忽略 | | 上市偏好 | 只扫描上市公司会遗漏产业链中的关键未上市玩家 | 必须搜索未上市公司，标注"未来IPO候选" | | 英文偏好 | AI对英文资料的处理能力更强，可能低估中国/亚洲市场玩家 | 必须同时搜索中英文信息源 |
产业链扫描中的反偏见措施：
使用 Task 工具启动后台 Agent，全面搜索该行业所有上市公司。
按产业链环节分类，每个环节一张表，包含所有扫描到的公司。 再按投资确定性分层：
对每个产业链环节的Tier 1和Tier 2公司，执行以下分析（Tier 3/4公司简要点评即可）：
用五类护城河评分（★1-5）：
| 护城河 | 强度 | 证据 | |--------|------|------| | 品牌/定价权 | | | | 转换成本 | | | | 网络效应 | | | | 规模效应 | | | | 技术/牌照壁垒 | | |
追问：10年后护城河还在吗？
用★1-5标注：
| 风险 | 概率 | 影响 | 应对策略 | |------|------|------|---------| | 投资逻辑链的某个环节被证伪 | | | | | 替代技术出现 | | | | | 政策/监管黑天鹅 | | | | | 需求周期性回调 | | | | | 估值泡沫破裂 | | | |
找到历史上类似的产业链投资主题，分析其最终结局：
按以下结构输出：
| 层级 | 仓位占比 | 标的 | 所属环节 | 核心逻辑 | |------|---------|------|---------|---------| | 核心仓位 | 占主题仓位50-60% | | | 最确定、护城河最宽 | | 卫星仓位 | 占主题仓位25-35% | | | 弹性较大、确定性稍低 | | 期权仓位 | 占主题仓位5-15% | | | 高风险高回报，可以归零 | | ETF替代 | 可替代以上全部 | | | 不想选股的"懒人方案" |
| 信号类型 | 具体条件 | |---------|---------| | 加仓信号 | | | 减仓信号 | | | 清仓信号 | |
根据投资逻辑链的确定性和风险程度，建议该主题占总仓位的上限百分比。
| 维度 | 结论 | 信心度 | |------|------|--------| | 投资逻辑链（验证程度） | | | | 最佳环节（段永平"对的生意"） | | | | 最宽护城河（巴菲特） | | | | 最大风险（芒格） | | | | 文明趋势定位（李录） | | | | 整体估值水平 | | |
用引用格式，模拟四位大师对该行业投资机会的点评。
~/[行业名]产业链投资研究报告.md报告写入后，执行数据抽检，通过方可发布：
bash# Step 1 — 提取抽检清单（15%随机抽样） python3 tools/report_audit.py extract \ --report <报告文件路径> # Step 2 — 对清单每项从可靠信源取数（参见 skills/financial-data.md） # Step 3 — 输出准出/打回判决 python3 tools/report_audit.py verdict \ --results '<填好的JSON>' \ --report <报告文件名>
【准出】 全部通过 → 报告可发布；【打回】 有不通过 → 修正后重审。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 41,476 | 5,314 | -87% | 1 | 1 | 0% | 6,278 | 3,254 | -48% | 0 | 0 | — |
case-01 | fail→fail | 48,581 | 8,522 | -82% | 1 | 1 | 0% | 6,284 | 3,522 | -44% | 0 | 0 | — |
case-02 | fail→fail | 41,631 | 8,857 | -79% | 1 | 1 | 0% | 6,299 | 3,592 | -43% | 0 | 0 | — |
case-04 | pass→fail | 30,177 | 14,457 | -52% | 1 | 1 | 0% | 6,210 | 3,215 | -48% | 0 | 0 | — |
case-05 | pass→fail | 94,191 | 8,869 | -91% | 1 | 1 | 0% | 5,442 | 3,178 | -42% | 0 | 0 | — |
case-06 | pass→fail | 30,239 | 5,126 | -83% | 1 | 1 | 0% | 4,355 | 3,163 | -27% | 0 | 0 | — |
case-07 | fail→fail | 15,315 | 6,251 | -59% | 1 | 1 | 0% | 2,231 | 3,170 | +42% | 0 | 0 | — |
case-08 | fail→fail | 31,798 | 7,506 | -76% | 1 | 1 | 0% | 237 | 3,382 | +1327% | 0 | 0 | — |
case-09 | fail→fail | 24,390 | 5,452 | -78% | 1 | 1 | 0% | 3,546 | 3,141 | -11% | 0 | 0 | — |
case-10 | pass→fail | 20,812 | 7,961 | -62% | 1 | 1 | 0% | 3,270 | 3,084 | -6% | 0 | 0 | — |
case-11 | pass→fail | 21,705 | 6,037 | -72% | 1 | 1 | 0% | 3,590 | 3,366 | -6% | 0 | 0 | — |
case-12 | fail→fail | 25,494 | 7,237 | -72% | 1 | 1 | 0% | 4,454 | 3,419 | -23% | 0 | 0 | — |
case-13 | pass→fail | 20,606 | 7,405 | -64% | 1 | 1 | 0% | 3,492 | 3,202 | -8% | 0 | 0 | — |
case-14 | fail→pass | 17,174 | 16,809 | -2% | 1 | 1 | 0% | 2,880 | 5,868 | +104% | 0 | 0 | — |
case-15 | fail→fail | 18,791 | 5,488 | -71% | 1 | 1 | 0% | 2,927 | 3,120 | +7% | 0 | 0 | — |
case-16 | fail→fail | 21,939 | 27,948 | +27% | 1 | 1 | 0% | 3,601 | 3,151 | -12% | 0 | 0 | — |
case-17 | fail→fail | 16,184 | 5,927 | -63% | 1 | 1 | 0% | 2,668 | 3,176 | +19% | 0 | 0 | — |
case-18 | fail→fail | 16,895 | 5,953 | -65% | 1 | 1 | 0% | 3,255 | 3,210 | -1% | 0 | 0 | — |
case-19 | fail→fail | 16,397 | 5,811 | -65% | 1 | 1 | 0% | 2,805 | 3,195 | +14% | 0 | 0 | — |
case-20 | fail→fail | 24,644 | 66,219 | +169% | 1 | 1 | 0% | 3,713 | 3,232 | -13% | 0 | 0 | — |
case-21 | fail→fail | 20,267 | 27,726 | +37% | 1 | 1 | 0% | 3,281 | 6,700 | +104% | 0 | 0 | — |
case-22 | fail→fail | 16,988 | 6,365 | -63% | 1 | 1 | 0% | 2,641 | 3,053 | +16% | 0 | 0 | — |
case-23 | fail→fail | 96,932 | 7,853 | -92% | 1 | 1 | 0% | 6,022 | 3,319 | -45% | 0 | 0 | — |
case-24 | fail→fail | 20,193 | 8,217 | -59% | 1 | 1 | 0% | 3,076 | 3,376 | +10% | 0 | 0 | — |
case-25 | pass→fail | 18,455 | 4,575 | -75% | 1 | 1 | 0% | 2,627 | 3,080 | +17% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted, and 2 counted toward the lift figure. The other 23 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -24 percentage points is the difference between those two pass rates over the 2 comparable cases. 11 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.