The Renowned [Youzi Skills] Has Joined Serenity and Other Judges

You all are very familiar with Youzi Skills – the most popular Skill in the stock analysis field. It has just been updated on June 8, with the number of judges reaching 65. This time, @aleabitoreddit and others have joined the panel.

This Skill is a deep individual stock analysis engine that lets 65 investment experts help you analyze the market. Currently, the 65-member panel across 9 domains is as follows:

1. AI Bottleneck/Chokepoint Hunter: Serenity (@aleabitoreddit)
2. Tech Leaders Faction: Jensen Huang (NVIDIA) · Elon Musk (Tesla) · Sam Altman (OpenAI) · Saylor (MSTR)
3. Quantitative Systems: Simons · Thorp · David Shaw · Asness (AQR)
4. A-Share Hot Money: Zhang Mengzhu · Zhao Laoge · Chaogu Yangjia · Foshan Wuyingjiao · Beijing Chaojia · Xin Duoduo, etc.
5. Chinese Value Investors: Duan Yongping · Zhang Kun · Zhu Shaoxing · Xie Zhiyu · Feng Liu · Deng Xiaofeng · Zhang Lei (Hillhouse)
6. Technical Trends: Livermore · Minervini · Darvas · Gann
7. Macro Hedging: Soros · Dalio · Howard Marks · Druckenmiller · Robertson · Burry (The Big Short) · Chanos (Short Hunter)
8. Growth Investing: Lynch · Cathie Wood · Thiel · Andreessen (a16z) · Gurley (Benchmark) · Naval · Gerstner (Altimeter) · Chamath
9. Classic Value: Buffett · Graham · Munger · Fisher · Templeton · Klarman

Repository: https://github.com/wbh604/UZI-Skill

I’m Nika, and I regularly share useful and interesting tools and projects in AI, US stocks, and Web3. If interested, feel free to follow. See you next time~

wbh604/UZI-Skill

Source: https://github.com/wbh604/UZI-Skill

UZI Skill

“51 investment experts help you analyze stocks – Buffett and Zhao Laoge finally sit at the same table.”

License: MIT
Python 3.9+ (https://python.org)
Claude Code (https://claude.com/product/claude-code)

Dimensions
Investors
Methods
Self-Review

A-Share / HK / US Stocks · Deep Individual Stock Analysis Engine · v3.8.0 Tier-1: Five Methods + Serenity Rigor (8 Penalties + Evidence Ladder + Supply Chain Layering) + DuPont/KDJ/OBV/Williams · v3.7.1 Added Serenity intro on homepage + --school H/I · v3.7.0 13 newly joined tech giants (52→65 judges)

Installation · Usage · Three Depth Levels · Hermes 🆕 · Panel · Serenity 🆕 · Institutional Methods · Self-Check Gate · Report Screenshots · FAQ · Group Communication · Contributors

中文 | English

🚀 Get Started in 30 Seconds

Just throw one sentence into any agent – it works after installation. Detailed installation steps are in Installation.

After installation, the four most common commands (say directly in any agent):

• /stock-deep-analyzer:analyze-stock 贵州茅台 ← Full 22-dimension × 65-judge analysis (5–8 min)
• /stock-deep-analyzer:quick-scan 002217 ← 30-second quick judgment
• /stock-deep-analyzer:scan-trap 002217 ← Scam stock screening
• /stock-deep-analyzer:dcf 600519 ← DCF valuation special

💡 Current latest stable version v3.3.2:

• v3.3.2 · 🆕 GitHub issues #50 + #51 hotfix (community-driven) · #50 institutional missing import svg_sparkline causing Stage 2 NameError · #51 XueQiu cubes_search.json endpoint offline → changed to query/v1/search/cube/stock.json (thanks @chenxiang-bj / @bilieebiliee1-design / @Kylin824)
• v3.3.1 · Hermes compatibility regression fix (v3.x refactoring period main branch missing Hermes adaptation; root cause of user errors) · main branch now directly supports hermes skills install
• v3.3.0 · Branch integration: cherry-picked segmental rendering layer into v3.2 architecture · Cleaned up 22 stale branches (retained only main + hermes-compat)
• v3.2.0 · assemble_report.py reduced from 2964 to 587 lines (-80%) · Split into 5 lib/report/*.py submodules
• v3.1.0 · run_real_test.py reduced from 2105 to 735 lines (-65%) · 1228 lines of pure functions moved to lib/pipeline/score_fns.py
• v3.0.0 · Pipeline architecture enabled by default (python run.py uses new path by default; UZI_LEGACY=1 for old path)

Two large files combined: 5069 → 1322 lines (-74%) · 332 tests all pass · End-to-end test 002217 resume generates report in 10s · All v2.x APIs 100% backward-compatible.

v2.15 series still retained: capital_flow universe cache (100x speedup) · school_scores grouped by school · Hybrid formula + polarization stretch.

Hermes users with old remnants may see errors · Reinstall once to fix (hermes skills uninstall then install 4 skills) · See INSTALL-HERMES.md.

💬 No Group!

I don’t know why the communication group got banned either… If I create a group, it gets banned. If you have ideas, add me directly. Only for the plugin itself or other interesting projects and quant topics, not individual stocks.

Too many people adding me; please include a note so I know what’s going on…

Acknowledgments

Learn AI at L Station! Thanks to Linux.do (https://linux.do/) community support.

What Is This?

In one sentence: input a stock, Claude becomes your personal analyst – runs data across 22 dimensions, calls 17 Wall Street analysis models, lets 51 investment experts with completely different styles each give a score, and finally produces a 600KB Bloomberg-style report.

/stock-deep-analyzer:analyze-stock 国盾量子

After 5–8 minutes, you’ll get:

• An HTML report – can be opened directly in a browser, self-contained, works offline
• A WeChat Moments vertical image – 1080×1920, ready to share
• A WeChat group battle card – 1920×1080
• A one-paragraph summary – copy and paste into the group

Why This Was Made

Previously, analyzing a stock required: East Money for fundamentals → Tonghuashun for K-line → Xueqiu for expert opinions → research reports for sell-side views → Excel for DCF → and still lose money after buying.

All these steps are essentially “collect information → analyze from multiple angles → draw a conclusion.” Why not have AI do it all?

Looking around, options were either GPT wrappers outputting three paragraphs of nonsense, or institutional terminals that are unaffordable. Anthropic released financial-services-plugins (https://github.com/anthropics/financial-services-plugins) with good methodology (DCF / Comps / LBO), but it’s purely US-stock focused and requires paid data sources.

So I built my own. Completely free data sources, zero API key, works directly for A-shares.

Installation

Regardless of which agent you use, just throw one sentence at it:

Claude Code

/plugin marketplace add wbh604/UZI-Skill
/plugin install stock-deep-analyzer@uzi-skill

After installation, say /stock-deep-analyzer:analyze-stock 贵州茅台.

⚠️ Must include stock-deep-analyzer: namespace prefix
After installing the plugin in Claude Code, all skills/commands start with stock-deep-analyzer:.
In some environments, short names (/analyze-stock) may not be automatically resolved – for safety, always use the full name:

• /stock-deep-analyzer:analyze-stock <symbol>
• /stock-deep-analyzer:quick-scan <symbol>
• /stock-deep-analyzer:scan-trap <symbol>
• /stock-deep-analyzer:dcf <symbol>
• /stock-deep-analyzer:ic-memo <symbol>
• /stock-deep-analyzer:investor-panel <symbol>
• /stock-deep-analyzer:trap-detector <symbol>
• /stock-deep-analyzer:deep-analysis <symbol>
• and all other 14 commands

Same for Cursor / Gemini CLI / Codex: always use full names with /stock-deep-analyzer: to avoid parsing failures.

Codex

Simply tell Codex:

Please follow the instructions at https://raw.githubusercontent.com/wbh604/UZI-Skill/main/.codex/INSTALL.md to install UZI-Skill, then help me do a deep analysis of 贵州茅台.

OpenClaw / Lobster

Tell Lobster:

Help me install the stock analysis skill from https://github.com/wbh604/UZI-Skill, then analyze 贵州茅台.

Cursor

/add-plugin stock-deep-analyzer

Then say “analyze 贵州茅台”.

Gemini CLI

bash gemini extensions install https://github.com/wbh604/UZI-Skill

OpenCode

Tell OpenCode:

Please follow https://raw.githubusercontent.com/wbh604/UZI-Skill/main/.opencode/INSTALL.md to install and analyze 贵州茅台.

Windsurf / Devin / Other Agents

Drop this sentence:

Clone https://github.com/wbh604/UZI-Skill, read AGENTS.md to learn how to use it, and help me do a deep analysis of 贵州茅台.

📱 Not at Your Computer?

Tell any agent:

Analyze 贵州茅台 using remote mode; generate a public link so I can view it on my phone.

The agent will automatically start with --remote using Cloudflare Tunnel and give you a https://xxx.trycloudflare.com link.

Usage

Full Deep Analysis (5–8 minutes)

/stock-deep-analyzer:analyze-stock 水晶光电
/stock-deep-analyzer:analyze-stock 002273
/stock-deep-analyzer:analyze-stock 00700.HK
/stock-deep-analyzer:analyze-stock AAPL

Specialized Commands

All require the /stock-deep-analyzer: prefix to guarantee execution.

🎯 Scoring Calibration (v2.11)

Users reported “Moutai scored 47” or “never above 65” – diagnosis found two overly strict formulas. v2.11 calibration:

Formula (unchanged): overall = fund_score × 0.6 + consensus × 0.4

Typical quality stock (e.g., Moutai) expectation:

• v2.9.1: fund=62 consensus=45 → overall 55 → watch priority
• v2.11: fund=62 consensus=50 → overall 57 → watch priority (but closer to “worth watching” boundary; when quality stock rally starts, easier to reach 65)

Combined effect of both tiers ~5–8 points. Real traps still < 35 → avoid, score discrimination actually improves.

Diagnostic field panel.json::consensus_formula.version = "v2.11 · (bullish + 0.6*neutral) / active" can be audited.

Regression test: tests/test_v2_11_scoring_calibration.py – 8 test cases.

Full calibration record in BUGS-LOG.md v2.11.0 section.

🎚️ Three Depth Levels (v2.10.3 New)

Give users the choice of analysis intensity – quick / normal / deep:

python run.py 600519 --depth lite   # ⚡ Quick mode (1–2 min)
python run.py 600519                 # 📊 Standard analysis (5–8 min) · default
python run.py 600519 --depth deep    # 🔬 Deep research (15–20 min)

Or via environment variable:

export UZI_DEPTH=lite   # or medium / deep
python run.py 600519

Summary of Level Differences

Auto-Downgrade Strategy

• First installation / .cache/_global empty → auto switch to lite (save first cold start time)
• Network pre-check 3+ domains unavailable → auto switch to lite (avoid freeze)
• Manual --depth always overrides auto decision

Practical Choices

Command Mapping (Implicit Tiers)

🎭 65-Member Panel

Not template phrases. Each person has their own quantitative rule set (236 rules total), and their recommendation must cite which specific rules were triggered.

Since v3.7.0, 13 newly added tech giants + an independent Group I Serenity (AI bottleneck/chokepoint hunter) cover 9 major schools:

v3.7.0 (2026-06) new 13 bold. Group H are tech CEOs with their own supply chain perspective; Group I Serenity is a single special judge (see below).

Example:

Buffett gives 水晶光电 62 points · Neutral
“Watch: moat 27/40 visible; but ROE 5-year low 6.7%, compliance rate only 0/5”
✅ Debt ratio 30% conservative · ❌ ROE 5-year low 6.7%

Jensen Huang gives a certain CPO optical module stock 100 points · Bullish
“In the AI compute chain · directly benefits from data center CapEx · gross margin ≥50% strong pricing power – it’s a beneficiary of Moore’s Law at the speed of light.”
✅ Strong correlation with AI compute demand · ✅ Deep CUDA/ecosystem binding

Klarman gives 水晶光电 0 points · Bearish
“Bearish core: no 30% margin of safety”

🧠 Group I · Serenity · AI Bottleneck/Chokepoint Hunter

Key figure: @aleabitoreddit (https://x.com/aleabitoreddit), a star overseas retail researcher that exploded on X (Twitter) in 2026.
Separated into its own group with its own score – because her style is highly concentrated, highly contrarian, unlike any institutional big name.

Who Is She?

• Self-described background: Former AI research scientist · Nature paper author · Former RISC-V Foundation member · Semiconductor/optical communications engineer
• Uses anime avatar, anonymous, no face reveal, doesn’t sell courses or copy-trade – all research free and public · 300K+ X followers
• Breakout win: Predicted the InP substrate bottleneck stock AXTI (12 → $70+, peak $115–140) about a year in advance. In Q1 2026, IntelliEPI CEO publicly confirmed “InP shortage is the bottleneck of the entire AI infrastructure.”

⚠️ Her identity and returns are self-reported/media-reported, not third-party audited; numbers from different sources contradict. This project only distills her methodology as an analytical perspective, not endorsing her actual track record. See docs/serenity-research-dossier.md (20+ sources with itemized records).

Her Role in UZI-Skill

Her AI industry chain bottleneck/chokepoint theory is turned into a quantifiable judge –
Don’t buy AI leaders (stocks like NVIDIA already fully priced), but go upstream along the supply chain to find that one “indispensable, most supply-constrained” small-cap gem in the second/third tier, and position before the market prices it in.

Leaders bought up → go upstream along supply chain → find the hardest-to-replace link → find the tightest-supply small cap in that link → position early

Core scoring logic: “position determines stance” – doesn’t look at cheap valuation or fast growth; only one variable: whether this company’s product is choking someone else in the current AI wave.

Deciding whether a link is a “chokepoint” looks at three things:

1. Hard to replace (how long to switch supplier/material/process – longer is better → 14_moat switching cost)
2. Tight supply (whether capacity can keep up with AI demand curve – less is better → 7_industry)
3. Not priced in (market still views it with old narratives like “cyclical stock / old semiconductor / niche material” → 5_chain + 15_events).

How to Run Serenity’s Perspective Alone

python run.py 300394.SZ --school I   # Only Serenity's "chokepoint" judgment
python run.py NVDA --school H         # Only Group H tech leaders (Jensen Huang/Musk/Altman/Saylor)

Six-step methodology + alpha 5 dimensions: see skills/deep-analysis/references/fin-methods/serenity-bottleneck.md;
Tone library + scoring rules: see skills/investor-panel/references/group-i-serenity.md.

📐 17 Institutional-Grade Methods

Porting methodologies from anthropics/financial-services-plugins (https://github.com/anthropics/financial-services-plugins), adapted to A-share parameters (rf=2.5% / ERP=6% / tax rate 25% / terminal growth g=2.5%):

Valuation Modeling

• DCF (WACC breakdown + two-stage FCF + Gordon Growth terminal + 5×5 sensitivity heatmap)
• Comps peer comparison (PE / PB / EV-EBITDA percentile + implied target price)
• Three-statement forecast (5-year IS / BS / CF linked)
• Quick LBO (PE fund perspective IRR cross-validation)
• M&A accretion/dilution model

Research Workflows

• Initiation report (JPM/GS/MS format · rating + target price + thesis + risks)
• Earnings beat/miss interpretation
• Catalyst calendar (real event extraction + future scheduling + impact grading)
• Investment logic tracking (5-pillar health)
• Morning note · quantitative screening · industry summary

Deep Decision-Making

• IC memo (8 chapters · Bull/Base/Bear three scenarios)
• Porter’s Five Forces + BCG matrix
• Due diligence checklist (5 workflows, 21 items · auto completion status)
• Unit economics · value creation plan · portfolio rebalancing

📸 What the Report Looks Like

All screenshots below are from real analysis of 水晶光电 (002273.SZ).

Composite Score + Key Conclusions

Bull-Bear Great Divide

Fisher 100 vs. Klarman 96, three rounds of arguments, each citing specific numbers.

65-Member Panel · Judgment Seats

Each person has a light – green bullish, red bearish, gray neutral.

Chat Room Mode

Judges speak in their own style, citing specific triggered rules.

DCF Valuation · 5×5 Sensitivity Heatmap

WACC 6.96% · Intrinsic Value ¥20.73 · Margin of Safety -28.6%, colors from dark green (undervalued) to dark red (overvalued).

IC Memo · Three-Scenario Returns

Bull ¥26.95 / Base ¥20.73 / Bear ¥14.51, each with probability and assumptions.

22-Dimension Deep Card

Each dimension has its own visualization – K-line candlestick / PE Band / radar chart / supply chain flow / thermometer / donut chart.

WeChat Moments Vertical Image · One-Click Share

🔧 Data Sources

All free, zero API key:

Multi-layer fallback chain – one source fails, automatically switches to next.

📱 6-Platform Social Hot Trends (v2.12 New)

Retail sentiment and scam themes often spread on Douyin/Xiaohongshu/Weibo first, which DuckDuckGo can’t reach. Since v2.12, the 17_sentiment dimension automatically checks:

• Weibo hot search · fetches weibo.com/ajax/side/hotSearch · 50 real-time trends
• Zhihu hot list · fetches zhihu.com/api/v3/feed/topstory/hot-list-web · 50 items
• Baidu hot search · fetches top.baidu.com/api/board · real-time list
• Douyin hot spots · fetches douyin.com/aweme/v1/web/hot/search/list/ · search hotspots
• Toutiao hot list · fetches toutiao.com/hot-event/hot-board/ · hot events
• Bilibili hot search · fetches s.search.bilibili.com/main/hotword · site-wide hot words

Stock name (including abbreviations, e.g., “贵州茅台” → “贵州”/“茅台”) if matched in trend titles → added to sentiment heat + record specific entries.

Data structure: synthesis’s 17_sentiment.data.hot_trend_mentions:

{
  "stock_name": "贵州茅台",
  "platforms_ok": 6,
  "total_hits": 3,
  "by_platform_count": {"weibo": 2, "zhihu": 1, ...},
  "mentions": {
    "weibo": [{"rank":3, "title":"茅台 1499 回归", ...}],
    ...
  }
}

Thanks: This module’s design references the hottrend service implementation from run-bigpig/jcp (https://github.com/run-bigpig/jcp) (Garlic Plate AI).

🔑 Optional: East Money Miaoxiang Skills API (v2.3 New)

In 2026, push2.eastmoney.com is often blocked by anti-scraping on mainland Chinese networks. If MX_APIKEY is set, UZI-Skill will prioritize the official NLP API:

• Chinese name correction: “北部港湾” → automatically recognized as “北部湾港(000582.SZ)”
• Market snapshot: Bypasses push2 to get latest price/market cap/PE/PB/industry

Configuration:

cp .env.example .env
# Edit .env to fill in MX_APIKEY (free to apply at: https://dl.dfcfs.com/m/itc4)

Without key, it falls back to Xueqiu/akshare chain; existing users are unaffected.

🔓 Data Sources Requiring Login (v2.7.1 New)

Some data sources added login authentication in 2026. UZI-Skill does not proactively pop up login windows (remains unattended). Users can enable as needed: