Recently, E Fund’s fund manager Zheng Xi has gained attention. He currently manages a total fund asset scale of approximately 24 billion yuan. His representative fund, E Fund Information Industry Mixed (001513), has achieved a cumulative return of nearly 980% since he took over in September 2016, with a tenure of nearly 10 years. As a result, someone in the community compiled all his publicly available quarterly reports, notes, and interview content from 2012 to 2026, distilling them into a proprietary Zheng Xi investment research Skill. I tried it myself, and this Skill can help you with the following:

1. Quickly look up Zheng Xi’s real historical views on a certain direction and their evolution
2. Conduct verbal-action comparisons: what he previously said he favored, whether his actual holdings actually included it, and when
3. Use his framework to score any fund and see the “Zheng Xi style” fit
4. Imitate his thinking for forward-looking judgments, or write market commentary in his tone

Most importantly, it does not fabricate. All conclusions either have original sources or clearly state they are derived from the framework.

You can use it directly in Claude Code or codeX.
Github: https://github.com/lyra81604/zhengxi-views

lyra81604/zhengxi-views

Source: https://github.com/lyra81604/zhengxi-views

🔎 Fund Manager · Zheng Xi Views Skill

Traceable · Original text as basis · Access to full-market funds · Cross-AI platform

“Check views · Learn methods · Make forward-looking judgments · Imitate tone to write commentary · Verbal-action comparisons · Compare and score full-market funds — let any AI understand E Fund’s Zheng Xi”

Many answers about “what a certain fund manager thinks about X” sound plausible but are actually fabricated by the model based on impressions — no such statement exists. This skill solves that problem: its foundation is the original text left by Zheng Xi (Deputy General Manager of E Fund Equity Investment Management Department, Fund Manager) himself. Every conclusion can be traced back to “which year, which article, what the original words were.”

It has three pillars, all from public and traceable sources:

• 📚 Original corpus (references/corpus/) — all of Zheng Xi’s public views from 2012–2026: investment operation analysis in periodic reports (quarterly/half-year/annual reports), fund manager notes, media interview reports, plus fund manager profile and lists of current and former funds managed.
• 🧭 Investment method (references/method.md) — a method framework distilled from the above corpus, with his own original words supporting each point. This allows the skill to extrapolate on topics the corpus hasn’t directly addressed, using Zheng Xi’s own methods, rather than saying “no such statement found.”
• 📊 Real fund data (references/fund_data/) — real data snapshots of all 8 funds he manages (4 current + 4 former): top 10 holdings each quarter, NAV/performance/scale/asset allocation/tenure returns. Plus a references/all_funds/ directory listing approximately 27,000 funds from the entire market for on-demand fetching to compare and score any fund.

⚠️ For research and learning assistance only, not investment advice.

🎯 What it can help you do

💬 Examples · See how it answers

Below are real outputs from the skill (slightly abridged).

1 Traceable Q&A — “What does Zheng Xi think about optical communication? When did he start being bullish?”

He clearly laid out the logic in a June 2026 interview with China Securities Journal:

“Global AI capital expenditure has already reached the trillion-dollar level. In the AI industry chain, data transmission is an important link, and optical communication is the key path for long-distance transmission. Combined with China’s global comparative advantage in optical communication, the market size for optical communication is expected to further increase.”

In terms of timing, he said his heavy position in “optical” can be traced back to Q2 2025. Cross-checking the holdings of the fund he manages, E Fund Information Industry (001513): starting from 2025Q2, optical communication stocks like Zhongji Innolight, Eoptolink, Yuantai Technology, and Optowide Technologies indeed entered the top ten collectively — words match actions.

2 Direction not covered in the corpus — “Use Zheng Xi’s thinking to look at innovative drugs?”

Zheng Xi has not publicly discussed innovative drugs. Below is a deduction based on his consistent investment method, not his personal opinion.

In all his public materials, “innovative drugs” are only mentioned in passing in quarterly report reviews, never elaborated. So applying his method:

• Gate 1: Does it have “demand created by the supply side”?: Innovative drugs create previously nonexistent demand through new targets/therapies, fitting his favored “tech-driven inflation” — this gate passes well;
• Gate 2: Where is China’s comparative advantage?: Original innovation originates overseas; China’s advantage lies more in CXO / out-licensing BD — the “water sellers” of the industry;
• Gate 3: Low ROE elasticity?: Unprofitable Biotech is a typical “pre-development low ROE,” matching his preference; but liquidity will filter out small-cap stocks with thin daily turnover. (Specific companies/valuation targets “need verification”; not investment advice.)

3 Zheng Xi framework scoring — “Use Zheng Xi’s criteria to score the China Merchants CSI White Wine ETF. Would he buy it?”

Scoring [China Merchants CSI White Wine Index (012414)] using Zheng Xi’s framework: 21 / 100 · Not a fit

In one sentence: This is a qualified sector index tool, but it almost doesn’t look like a “fund Zheng Xi would buy.” Low score ≠ bad fund, just that the style is almost opposite to his.

Dimension	Score	One-sentence basis
Hot direction / inflation attribute	4 / 25	Top 10 are all high-end white wine, currently in a down cycle of destocking, “being priced down” not “pricing up”
Low ROE elasticity	6 / 20	All are white wine blue chips with persistently high ROE, not his desired “low, waiting for recovery”
Global perspective / China comparative advantage	2 / 15	Pure domestic consumption, unrelated to global technology cycles
Liquidity	7 / 10	All top holdings are large-cap blue chips with best liquidity (only bright spot)
Concentration and cycle stitching	2 / 15	Concentration 85.8% but passive lying flat, ~5% turnover proxy in recent 5 quarters, opposite to his “cycle stitching/high turnover”
Performance and drawdown confirmation	0 / 15	-23.9% in 1 year, max drawdown -63.2%, significantly underperformed

Would he buy it? Almost certainly not — a passively managed single traditional sector index fund is completely mismatched with his approach of “top-down looking for inflation, favoring tech growth, buying low ROE elasticity, dynamic cycle stitching.”

4 Full-market comparison — “Compare Zheng Xi’s Information Industry with Ge Lan’s China Euro Medical”

	Zheng Xi · Information Industry 001513	Ge Lan · China Euro Medical 003095
Sector	Tech (optical communication, computing power, semiconductors)	Pharma (CXO, innovative drugs)
Style	Top-down, industry chain rotation, high turnover	Core assets long hold, low turnover
1-year return	+290%	-5.4%
Max drawdown	-46.3%	-63.9%

In one sentence: Both are high-conviction single-track players, but Zheng Xi turned elasticity into gains by “riding the current hot sector + dynamic rotation,” while Ge Lan “long holding core assets” fully bore deeper drawdowns during the pharma down cycle. (Data from public quarterly snapshots as of June 2026; not investment advice.)

⚖️ Differences from “investment framework/methodology” skills

🧩 Is it an Agent Skill?

Yes. It conforms to the Agent Skill specification: root directory includes SKILL.md (YAML frontmatter with name + description) + progressively loaded references/ and scripts/ resources, which can be validated with official tools and packaged as a .skill package.

• Claude Code / Claude.ai natively support Agent Skill — place in the skills directory and it’s automatically loaded.
• Tencent WorkBuddy uses a skill.yml manifest to organize Skills. This repository includes one; place it in WorkBuddy’s skills/ directory (tested and working).
• In ChatGPT / Gemini / Cursor and other tools without an “auto-skill loader,” use “custom instructions + knowledge files” to replicate — see Using in other AI tools.

🗂️ Directory Structure

zhengxi-views/
├── SKILL.md
├── skill.yml                    # Tencent WorkBuddy manifest (can be ignored for Claude usage)
├── WORKBUDDY部署.md
├── README.md
├── references/
│   ├── method.md
│   ├── scorecard.md
│   ├── corpus_index.json
│   ├── corpus/
│   │   ├── 定期报告/             # Periodic reports
│   │   ├── 基金经理手记/         # Fund manager notes
│   │   ├── 媒体报道/             # Media reports
│   │   ├── 简介.md               # Profile
│   │   ├── 管理基金_在任.md      # Funds currently managed
│   │   └── 管理基金_曾任.md      # Funds previously managed
│   ├── fund_data/
│   │   ├── _index.md
│   │   ├── 001513_易方达信息产业混合/  # E Fund Information Industry Mixed
│   │   └── ... (8 funds total, each with quarterly holdings.md and NAV performance scale.md)
│   └── all_funds/
│       └── fund_list.json
└── scripts/
    ├── search_corpus.py
    ├── build_index.py
    ├── fetch_fund_data.py
    ├── build_fund_list.py
    ├── fund_lookup.py
    ├── fetch_any_fund.py
    └── score_fund.py

⚙️ Installation

Claude Code (native, recommended)

macOS / Linux:

mkdir -p "$HOME/.claude/skills/zhengxi-views"
cp -R SKILL.md README.md references scripts "$HOME/.claude/skills/zhengxi-views"/

Windows PowerShell:

$dst = "$HOME\.claude\skills\zhengxi-views"
New-Item -ItemType Directory -Force $dst | Out-Null
Copy-Item -Recurse SKILL.md,README.md,references,scripts $dst

After installation, fully restart Claude Code and ask “What does Zheng Xi think about optical communication?” to trigger it.

Dependencies (only needed for “fetch fund data / scoring”)

Pure corpus search does not require third-party libraries. To use fetch_* / score_fund for full-market data, first install:

pip install -r requirements.txt  # requests / beautifulsoup4 / lxml

Reduce per-step confirmation (Claude Code, optional)

The skill runs python scripts. By default, Claude Code asks for confirmation each time. To skip confirmation, add a python allow rule in ~/.claude/settings.json under permissions.allow:

{
  "permissions": {
    "allow": ["Bash(python:*)", "Bash(python3:*)"]
  }
}

(This will allow all python commands. Recommended only on personal machines. Can be removed anytime. The script calls are designed as single commands with no cd/redirection to avoid security confirmations.)

🌐 Using in other AI tools

This project is an Agent Skill, natively supported by Claude series; other tools replicate it with “instructions + knowledge files.” The available features depend on whether the tool can run Python and access the internet:

Tencent WorkBuddy runs locally and can run scripts with internet access, so it can use all features (including real-time fetch). It uses skill.yml as manifest; this repo includes one — see Tencent WorkBuddy. ⚠️ = sandbox usually has no external network, so real-time fetch likely fails. However, you can first run fetch_any_fund.py locally to fetch the target funds, then upload the generated data files together, and offline scoring will work normally.

General approach (any AI supporting custom instructions + file uploads):

• Instructions: Paste the body of SKILL.md into “System Prompt / Custom Instructions.” The section “How to run scripts” is specific to Claude Code — can be deleted or changed to “Run uploaded scripts with code interpreter.” Be sure to keep behavioral constraints (traceability, no fabrication, separate extrapolation from original words, bold statement if outside corpus).
• Knowledge: Upload all content from references/ (method.md, scorecard.md, corpus/, fund_data/, all_funds/fund_list.json). For script use, also upload scripts/.

Tencent WorkBuddy

WorkBuddy is a locally running AI workbench. Custom Skills follow the skill.yml manifest + implementation files (scripts) + README convention. This repo already includes skill.yml, ready to use:

1. Place the entire folder into WorkBuddy’s skills/ directory, import and enable via skill.yml (tested and working).
2. First pip install -r requirements.txt to install script dependencies. Because it runs locally, all features including real-time fetch/scoring are available.
3. Trigger examples: “What does Zheng Xi think about optical communication?” or “Use Zheng Xi’s criteria to score the China Merchants CSI White Wine ETF.”

skill.yml fields follow a general convention. If WorkBuddy prompts field mismatches, adjust according to its schema (behavior logic is all in SKILL.md). Complete steps and alternative plans are in WORKBUDDY部署.md.

Cursor (IDE, most feature-complete, close to Claude Code)

1. Place the entire zhengxi-views/ folder into the workspace.
2. Create .cursor/rules/zhengxi.mdc and paste the body of SKILL.md as a Rule.
3. Cursor’s Agent has a local terminal, can run python + internet access — search, real-time fetch, scoring all work. Usage is the same as Claude Code.

ChatGPT (Custom GPT)

1. Create a new GPT (My GPTs → Create).
2. Instructions: Paste the body of SKILL.md (adjust script-related wording per the “general approach” above).
3. Knowledge: Upload method.md, scorecard.md, corpus/ (can be zipped), fund_data/, all_funds/fund_list.json, and scripts/*.py.
4. Enable Code Interpreter (Data Analysis).
   • Available: traceable Q&A, methods, stylistic commentary, Zheng Xi fund data, script-based search/scoring on uploaded data.
   • Limited: real-time fetch of full-market funds (Code Interpreter has no external network) — instead, fetch locally first and then upload.

Gemini (Gem / Google AI Studio)

1. Create a new Gem.
2. Paste the body of SKILL.md into the instructions field; upload method.md, scorecard.md, corpus/, fund_data/ as knowledge.
3. Primarily supports read features (Q&A / method / commentary / comparison of packaged data). No code + internet, so real-time fetch and script scoring should be done locally first, then bring results to ask questions.

Claude.ai (Web / Desktop)

• If your account supports uploading Skill, directly upload the packaged .skill file; or create a new Project and upload SKILL.md with references/ and scripts/ as project knowledge.
• The web sandbox may have no external network, so real-time fetch may be limited; built-in corpus and Zheng Xi fund data work normally.

🛠️ Quick Script Reference

search_corpus.py "optical communication"       Search corpus, return matching paragraphs + source (new→old)
search_corpus.py "ROE" "elasticity" --any      Match any keyword
fund_lookup.py "China Euro Medical"            Search full market by name/code/type for fund code
fetch_any_fund.py 003095                       On-demand fetch fund holdings/NAV/performance to cache
score_fund.py "China Merchants CSI White Wine" One-command entry for Zheng Xi framework scoring (find code + prepare data + compute indicators)
build_index.py / build_fund_list.py            Rebuild after corpus / full-market list update

In Claude Code / Cursor, the AI will call these scripts itself; you usually do not need to run them manually.

✅ Data Sources and Authenticity

Corpus, investment method evidence, and fund data all come from publicly disclosed content by Zheng Xi and the funds he manages — periodic reports on the E Fund website, fund manager notes, media reports, and public fund data from Tiantian Fund. All are real materials, traceable to original text. The skill only retrieves and cites, does not fabricate.

🚧 Limits

For research and learning assistance, not investment advice. Does not predict price movements, give buy/sell instructions, or promise returns. Citations must be faithful to the original text. The biggest taboo is fabricating something he never said or rewriting original text to pass it off as a quote — when in doubt, go back to the corpus, or honestly say “no record found.”

📄 License

MIT