Guide · Agent skill
How to use the UZI Deep Stock Analysis agent skill (deep-analysis)
Updated · written with AI from the project's own documentation and checked against our catalogue data
deep-analysis is an agent skill from the UZI-Skill project that turns an AI agent into a stock analyst for a single listed company on A-share, Hong Kong or US markets. This guide explains what the skill makes the agent do, how to install it, how to trigger it with ordinary requests, and what to check for safety before running it. By the end you will be able to add the skill, ask for a deep analysis, and read the output with the right level of caution.
UZI Deep Stock Analysiswbh604/UZI-Skill · MIT Download the skill Quick start
- 1Download the ZIP. Get the deep-analysis skill ZIP from the directory page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. Open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Install Python dependencies. Install Python 3 and the akshare and fetcher libraries the scripts import.
- 5Make a natural request. Ask for a deep analysis of one A-share, Hong Kong or US stock by name or code.
- 6Review the report. Check the conflicts, verify key numbers and treat the scores as model opinions.
What is an agent skill, and what does deep-analysis do?
An agent skill is a folder containing a SKILL.md file that tells a compatible AI agent when to use it and how to work through a task. The deep-analysis skill is the core workflow for analysing one stock. Its description lists triggers such as 'deep analysis', 'full analysis', 'is it worth buying', 'DCF', 'initial coverage' and 'investment committee memo'.
When triggered, the skill tells the agent to gather data through scripts across about 22 data dimensions, run institutional-style valuation models, have 65 investor personas score the stock by rule, and then write its own judgement into a Bloomberg-style HTML report. The source says the agent must write the final judgement and narrative itself; the scripts do the arithmetic.
- Data gathering: about 22 fetchers covering basic info, financials and more
- Valuation models: DCF, Comps, LBO, 3-Statement and Merger, among others
- Panel review: 65 investor personas scored by rule, each citing specific rules
- Trap detection: a 'pig-butchering scam' check
- Output: an HTML report and a shareable summary card
Install the deep-analysis skill
There are two routes. You can download the ZIP from the directory page, unzip it into your personal skills folder or a project folder, or upload it in the Claude apps.
For a personal install, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill there.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
Prepare Python and the data libraries
The skill needs Python 3 and the run.py entry point. It also depends on akshare and other data-fetching libraries used by its fetchers. The source does not list the full set of libraries, so check the fetcher files and install what they import before your first run.
The skill also needs network access to market data sources, and an agent that reads SKILL.md, such as Claude Code.
- Python 3 with run.py
- akshare and other fetcher libraries (not fully listed in the provided files)
- Network access to market data sources
- Claude Code or another SKILL.md-compatible agent
Trigger the skill with a natural request
You do not need a special command. Ask in plain language about one company, using a name or a code. The skill is written to trigger on phrases like 'deep analysis', 'full analysis', 'help me look at', 'is it worth buying', 'DCF' or 'investment committee memo'.
If the name is ambiguous, the skill tells the agent to stop and ask you to choose among candidates rather than guess. If you enter an ETF, fund or convertible bond, the skill redirects you, since it only covers individual stocks.
- Full look: 'Do a deep analysis of 002273.SZ'
- Valuation focus: 'Run a DCF and comps on this stock'
- Decision memo: 'Write an investment committee memo for this company'
- Ambiguous name: expect a list of candidates to choose from
Read the output with care
The skill asks the agent to show conflicts rather than smooth them over. For example, if DCF and Comps disagree, the report should say so, and wide disagreement among personas is presented as information in itself.
Treat every score and comment as a model-generated opinion. The bundled disclaimer states that persona comments are algorithmic simulations and that the named investors have not seen them. Use the report to check your own reasoning and verify the numbers against primary sources.
- Conflicts between models are kept in the report
- Persona comments are simulations, not real investors' views
- Data may lag, contain errors or be incomplete
Safety notes: scripts and network access
This skill runs scripts and fetches live market data over the network. Read the scripts before running them, especially anything that makes network requests, and run it in a folder you control. Do not use it if you cannot review code.
The disclaimer also warns that a recommendation from an AI is not proof a stock will rise, and that if someone is pushing you to buy a stock with claims of insider tips or 'AI says buy', you should run the trap detection step first.
- Inspect scripts before first run
- Expect network calls to third-party data sources
- Trap detection is heuristic and cannot cover every scam method
What people use it for
- Getting a structured second opinion on a specific A-share, Hong Kong or US stock
- Seeing how a multi-method equity research workflow is organised
- Studying how an agent skill chains scripts, personas and report generation
- Preparing an investment committee style memo for discussion
Frequently asked questions
How do I install the deep-analysis skill?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings, Capabilities, Skills.
What does the deep-analysis skill need to run?
It needs Python 3 with the run.py entry point, akshare and other data-fetching libraries, network access to market data sources, and an agent that reads SKILL.md such as Claude Code. The full library list is not given in the provided files.
Is deep-analysis free to use?
The source states the skill is licensed under MIT. It does not give pricing for data sources, so check the licence and any third-party services you connect.
Does it work for ETFs or funds?
No. The skill covers individual stocks only. For ETFs, it is designed to show top holdings and ask you to pick a constituent stock. For LOF funds or convertible bonds, it redirects you to other tools.
Is it safe to run?
It runs scripts and fetches live data over the network, so the source advises inspecting the code before use. Only run it if you can review the scripts and understand the network calls.
How does it compare with other stock analysis tools?
The source does not compare it with other tools, so this guide cannot make a claim. Its distinctive features are the multi-method valuation set, the persona panel and the trap detection step.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use deep-analysis if you want a structured, multi-method second opinion on one listed stock and you can review its scripts and check its numbers. Pick something else if you need trading decisions without verification, want ETF or fund analysis, or cannot inspect code that fetches live data.
Download the skill