Codex adapter note
This skill is generated from skills/industry-research.md so Claude Code and Codex users share one canonical workflow.
- Treat
$ARGUMENTSas the user's request in the current Codex thread. - When the source mentions Claude-only surfaces such as Task, Agent, WebSearch, Bash, Read, or Write, use the closest Codex capability available in this session: subagents when available, web search when needed, shell commands for local tools, and normal file edits for workspace files.
- Use shared project tools from
tools/in this repository. Prefer running commands from the repository root with paths likepython3 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. - Before starting research, run the
datecommand 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. - Preserve the research quality rules from
AGENTS.md: cross-check financial data, use exact arithmetic tools for valuation/math, and clearly label uncertainty and source gaps.
Conduct systematic industry-chain investment research on the $ARGUMENTS industry.
Research objectives
Starting from an investment theme/logic chain, complete:
- Verify every link in the investment logic chain
- Draw a complete industry-chain panorama
- Scan all listed companies globally (A-shares/Hong Kong stocks/US stocks/international)
- Execute the Four Masters framework analysis on leading companies in each segmented link
- Output industry-level portfolio allocation recommendations
Step 1: Investment logic chain construction and verification
1.1 Draw the logic chain
Use arrow links to express the causal relationship from "underlying trend" to "beneficiary targets", for example:
底层趋势 A
→ 导致需求 B
→ 创造瓶颈/刚需 C
→ 受益产业链 D
1.2 Verify link by link
Question every arrow in the logic chain and look for evidence:
| Link | Core assumption | Verification method | Data source |
|---|---|---|---|
| A→B | Search industry data/forecasts | ||
| B→C | Search supply-demand analysis | ||
| C→D | Search actual cases/signed contracts |
1.3 Look for "verification events that have already occurred"
List thereal business events already signed/implementedthat support this logic chain (rather than forecasts), such as large companies' procurement agreements, policy documents, industry reports, etc.
Step 2: Industry-chain panorama drawing
2.1 Draw the industry-chain structure
Break the industry down into upstream → midstream → downstream → auxiliary links, for example:
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）
2.2 Identify the "business characteristics" of each link
Label each link with:
| Link | Business model | Gross margin range | Competitive landscape | Barrier type | Cyclicality |
|---|---|---|---|---|---|
| Selling resources/selling equipment/selling services/collecting rent | Monopoly/oligopoly/full competition | Resources/license/technology/scale | Strong/medium/weak |
2.3 Mark "chokepoint links"
Identify the links in the industry chain with the tightest supply, the hardest substitution, and the highest profit margins—these are often where the best investment targets are.
AI research bias awareness: Special pitfalls of industry research
In industry research, AI data bias is amplified in unique ways:
Industry-level bias：
| Bias type | Manifestation | Response |
|---|---|---|
| Preference for mature industries | Traditional industries (banking/energy/consumer) have abundant materials, and AI analysis appears "more certain" | Certainty comes from the business model, not from the number of research reports |
| Underestimation of emerging industries | New industries (AI applications/synthetic biology, etc.) have little material, and AI analysis tends to be conservative | Judge industry value using "endgame thinking" rather than "current data" |
| Preference for leaders | Large companies have far more material than small companies, and AI naturally tends to recommend leaders | Small companies may have a better risk-return ratio; do not ignore them just because the AI analysis is short |
| Listed-company preference | Scanning only listed companies will miss key unlisted players in the industry chain | Must search for unlisted companies and label them as "future IPO candidates" |
| English preference | AI has stronger processing capability for English materials and may underestimate players in the China/Asia market | Must search both Chinese and English information sources simultaneously |
Anti-bias measures in industry chain scanning：
- For each segment, not only list "companies AI can easily find," but also proactively search for "obscure but potentially high-quality targets"
- For small-cap companies with scarce information, do not lower the recommendation level just because the analysis is short—judge by core questions (business essence, moat, management) rather than report length
- In the final report, label each company's "information sufficiency" (Grade A/B/C) so readers know the reliability of the AI analysis
Step 3: Global listed company scan
Use the Task tool to launch a background Agent to comprehensively search all listed companies in the industry.
Search checklist
- U.S. stocks (NYSE/NASDAQ/NYSE American) related companies
- A-shares (Shanghai/Shenzhen) related companies
- Hong Kong stocks related companies
- Other international markets (Japan/South Korea/Europe/Australia, etc.)
- Industry ETFs
- Key unlisted companies (possible future IPOs)
For each company, collect
- Company name (Chinese and English)
- Stock ticker and exchange
- Market cap (approximate)
- One-sentence description (position and role in the industry chain)
- Whether it is a pure-play target (pure nuclear power vs. diversified with nuclear power business)
- Industry chain segment
Output format
Classify by industry chain segment, with one table per segment, including all companies scanned. Then layer by investment certainty:
- Tier 1: Large market cap, pure-play target, industry leader
- Tier 2: Mid market cap, pure-play or high proportion, segment leader
- Tier 3: Small market cap, development stage, high risk and high elasticity
- Tier 4: Large enterprises with related businesses among diversified companies
Step 4: Four Masters analysis of leading companies in each segment
For each industry chain segment'sTier 1 and Tier 2 companies, perform the following analysis (Tier 3/4 companies only need brief comments):
4.1 Business essence (Duan Yongping)
- Define in one sentence what this company does in the industry chain
- Revenue structure and growth rate
- Gross margin/net margin levels and trends
- Cash flow characteristics
- Follow-up question: Is this a good business? Why?: Is this a good business? Why?
4.2 Moat (Buffett)
Score using five types of moats (★1-5):
| Moat | Strength | Evidence |
|---|---|---|
| Brand/pricing power | ||
| Switching costs | ||
| Network effects | ||
| Scale effects | ||
| Technology/license barriers |
Follow-up question: Will the moat still exist in 10 years?: Will the moat still exist in 10 years?
4.3 Risk (Munger)
- How is this company most likely to fail?
- What is it worth in the worst-case scenario?
- Why don't smart people buy it?
4.4 Management (Duan Yongping + Buffett)
- Who is the CEO/founder? Key decision records
- Shareholding ratio and alignment of interests
- Brief assessment (Grade A/B/C)
4.5 Valuation snapshot
- Current PE/PS/EV/EBITDA
- Comparison with competitors in the same segment
- Brief assessment: expensive/fair/cheap
4.6 Recommendation level
Mark with ★1-5:
- ★★★★★ = Core position candidate
- ★★★★☆ = Satellite position candidate
- ★★★☆☆ = Watchlist
- ★★☆☆☆ = High-risk option
- ★☆☆☆☆ = Not recommended
Step 5: Industry-level risk assessment (Munger's "checklist")
5.1 Systemic risk checklist
| Risk | Probability | Impact | Response strategy |
|---|---|---|---|
| A certain link in the investment logic chain is falsified | |||
| Emergence of substitute technology | |||
| Policy/regulatory black swan | |||
| Cyclical demand downturn | |||
| Valuation bubble burst |
5.2 Historical analogy
Find similar industry chain investment themes in history and analyze their final outcomes:
- What is the analogous industry?
- Who was the final winner? (Upstream/midstream/downstream?)
- Did most investors make money or lose money?
- What are the implications for the current industry?
5.3 Bias self-check
- Narrative bias: Is the story too perfect?
- Anchoring effect: Are you anchored by recent gains?
- Herd effect: Is it because "everyone is buying"?
Step 6: Civilizational trend judgment (Li Lu framework)
- Is the underlying trend this industry relies on a "civilization-level paradigm shift" or a "phase-specific boom"?
- What is the closest technological revolution analogy in history?
- In 10-20 years, what is the endgame of this industry?
- In the industry chain, which segment is most likely to have a "winner takes all"?
- Which segment is most likely to be disrupted?
Step 7: Portfolio allocation recommendations
7.1 Recommended portfolio
Output according to the following structure:
| Tier | Position proportion | Target | Segment | Core logic |
|---|---|---|---|---|
| Core position | 50-60% of theme position | Most certain, widest moat | ||
| Satellite position | 25-35% of theme position | Greater elasticity, slightly lower certainty | ||
| Option position | 5-15% of theme position | High risk, high return, can go to zero | ||
| ETF alternative | Can replace all of the above | A "lazy person's solution" for those who do not want to pick stocks |
7.2 Buy/sell signals
| Signal type | Specific conditions |
|---|---|
| Add-position signal | |
| Reduce-position signal | |
| Exit signal |
7.3 Recommended theme position cap
Based on the certainty of the investment logic chain and the degree of risk, recommend the maximum percentage of total position for this theme.
Step 8: Comprehensive decision memorandum
Industry Overall Assessment Table
| Dimension | Conclusion | Confidence Level |
|---|---|---|
| Investment Logic Chain (Degree of Verification) | ||
| Best Segment (Duan Yongping's "Right Business") | ||
| Widest Moat (Buffett) | ||
| Biggest Risk (Munger) | ||
| Civilization Trend Positioning (Li Lu) | ||
| Overall Valuation Level |
Simulated Commentary by the Four Masters
Use quote format to simulate the four masters' commentary on the investment opportunities in this industry.
Output Requirements
- All analysis must be supported by data, with data sources attached
- Use Markdown tables to present key data
- Represent the industry chain panorama with a text diagram in a code block
- Analyze at least 2-3 leading companies for each segment
- The global company scan should be as complete as possible (A-shares/Hong Kong stocks/US stocks/international)
- Finally write the complete report into
~/[行业名]产业链投资研究报告.md - The conclusion must be clear, giving specific targets, position sizes, and price range recommendations
- At the end of each analysis module, include the corresponding master's "follow-up question"
Data Spot Check (Exit Process)
After the report is written, perform a data spot check, and only release it if it passes:
# 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>
[Pass] All pass → the report can be published;[Reject] If any fail → revise and re-review.