Agentic ai β€’ β€’ 5 min read

πŸ›  Claude Skills - Hands-On: Build a Stock Analyzer Expert with Claude Skills

In general, a Claude agent can do many things β€” but if you want it to be truly expert in a specific domain, **Claude Skills** is the answer. Skills are more organized, easier to execute, and produce more consistent results than open-ended prompting.

Agentic AI Claude Skills Claude Code AI Agents LLM Gen AI Ollama Local AI Models

A practical, step-by-step workshop

πŸ“– Read the intro first: Claude Skills β€” Introduction & Overview

Table of Contents

  1. What We’re Building
  2. What the Skill Must Achieve
  3. How Claude Achieves Goals Using Skills
  4. The Skill Package Layout
  5. SKILL.md β€” Where Everything Starts
  6. helper-script.py β€” All Functions in One Place
  7. Installing the Skill in Claude Code
  8. Run It β€” Analyze TESLA Stock

1. 🎯 What We’re Building

In general, a Claude agent can do many things β€” but if you want it to be truly expert in a specific domain, Claude Skills is the answer. Skills are more organized, easier to execute, and produce more consistent results than open-ended prompting.

In this workshop we create a real, working Stock Analyzer Skill β€” a Claude agent that:

  • Fetches live market data from Yahoo Finance
  • Computes professional technical indicators
  • Generates a comprehensive interactive HTML report

All triggered by a single natural-language query.

πŸ’‘ Skills make Claude deterministic. Instead of reasoning from scratch every time, Claude follows your pre-tested scripts and produces the same high-quality output on every run.

2. πŸ† What the Skill Must Achieve

Our Stock Analyzer Skill has three concrete, measurable goals. Everything we build serves one of these.

# Goal Description
πŸ“₯ 01 Fetch Historical Data Read real OHLCV price data from the Yahoo Finance API for any stock ticker and time period
πŸ“Š 02 Technical Analysis Compute RSI, MACD, Bollinger Bands, SMA 20/50/200, EMA 12/26 β€” automatically on every run
πŸ“„ 03 Generate Report Produce a professional HTML report with interactive Plotly charts and fundamental metrics

3. βš™οΈ How Claude Achieves Goals Using Skills

Every Claude Skill’s instructions must be defined in a SKILL.md file β€” the brain of the skill. This is where you define the goal and point to the required scripts, helpers, and references so Claude knows exactly what to do.

πŸ’¬ 1. User sends a natural query
        "Analyze TESLA stock using stock-analyzer"
                    ↓
πŸ” 2. Claude matches query to the Skill
        Scans name + description in SKILL.md
                    ↓
🧠 3. Claude reads SKILL.md and follows instructions
        Loads full instructions into context
        Knows which scripts, templates, references to use
                    ↓
🐍 4. helper-script.py executes
        Fetches data β†’ computes indicators β†’ renders report
                    ↓
βœ… 5. Output delivered
        HTML report saved Β· key findings summarized in terminal

4. πŸ“ The Skill Package Layout

The skill is a folder. Claude Code discovers it inside .claude/skills/ in your working directory.

claude-workspace/              ← open terminal from here
└── .claude/                   ← Claude Code config folder
    └── skills/                ← all skill packages live here
        └── stock-analyzer/    ← your skill package
            β”œβ”€β”€ SKILL.md       ← brain of the skill (required)
            β”œβ”€β”€ helper-script.py  ← analysis + report functions
            β”œβ”€β”€ templates/     ← report.html, email.txt
            └── resources/     ← reference-data.json

What each file does

File Role Description
πŸ“‹ SKILL.md Brain Β· Required YAML metadata for discovery + full instructions for execution
🐍 helper-script.py Core Engine All functions: analyze_stock(), generate_report(), generate_email(), all indicator calculations
πŸ“ templates/ Output Blueprints report.html with `` Β· email.txt summary template
πŸ“ resources/ Reference Data reference-data.json β€” RSI thresholds, sector P/E benchmarks, metric definitions

5. 🧠 SKILL.md β€” Where Everything Starts

SKILL.md has two parts:

  • YAML frontmatter β€” lightweight metadata Claude always reads (always in memory)
  • Full instruction body β€” loaded only when the skill triggers
---
name: stock-analyzer
description: Comprehensive stock market analysis using Yahoo Finance data.
  Use this skill when users request stock analysis, historical price
  data, financial metrics, technical indicators, or investment reports
  for any stock symbol. Triggers include "analyze AAPL stock",
  "get Tesla financial data for last 6 months", "create stock report
  for MSFT", "compare NVDA performance over 1 year", or any mention
  of stock tickers with time periods. Generates detailed HTML reports
  with interactive charts, technical analysis, and fundamental metrics.
---

# Full instructions below β€” loaded when skill triggers
# Tells Claude: which script to run, what outputs to generate,
# which templates/resources to use, and how to interpret results.

✦ Write the description like a trigger condition. Include real example phrases users would actually say. Claude matches these patterns to decide whether to fire the skill.

Field Limit Purpose
name 64 chars Unique identifier Claude sees during discovery
description 1,024 chars The trigger condition β€” make every word count
Full instructions < 5k tokens The detailed playbook for Claude
Linked files On-demand Scripts, templates, data loaded only if needed

6. 🐍 helper-script.py β€” All Functions in One Place

helper-script.py is the single Python file that does all the heavy lifting. It contains every function the skill needs β€” pre-tested and production-ready.

Core Functions

Function What it does
πŸ“₯ analyze_stock(symbol, period) Fetches Yahoo Finance data + computes all indicators. Returns a rich dict.
πŸ“Š calculate_rsi(data, period=14) 14-period RSI β€” overbought >70, oversold <30
πŸ“ˆ calculate_macd(data) MACD line, signal line, histogram (12/26/9)
〰️ calculate_bollinger_bands(data) Upper/mid/lower bands (20-period, Β±2 std dev)
πŸ“„ generate_report(data, path) Loads report.html, fills ``, writes self-contained HTML
βœ‰οΈ generate_email(data, recipient) Renders email.txt template, returns formatted string

RSI Calculation

def calculate_rsi(data, period=14):
    """Relative Strength Index (default 14-period)."""
    delta = data['Close'].diff()
    gain  = (delta.where(delta > 0, 0)).rolling(window=period).mean()
    loss  = (-delta.where(delta < 0, 0)).rolling(window=period).mean()
    rs    = gain / loss
    return 100 - (100 / (1 + rs))   # overbought >70 Β· oversold <30

Stock Analysis β€” What Gets Returned

def analyze_stock(symbol, period='1y'):
    stock = yf.Ticker(symbol)
    hist  = stock.history(period=period)

    # Technical indicators computed automatically
    hist['SMA_20']  = calculate_sma(hist, 20)
    hist['SMA_50']  = calculate_sma(hist, 50)
    hist['SMA_200'] = calculate_sma(hist, 200)
    hist['RSI']     = calculate_rsi(hist)

    macd, signal, hist_vals = calculate_macd(hist)
    upper, mid, lower       = calculate_bollinger_bands(hist)

    # Returns full analysis dict
    return {
        'company_info':         ...,  # name, sector, employees, summary
        'current_price':        ...,  # price, change, change_percent
        'key_metrics':          ...,  # PE, PB, ROE, margins, debt/equity
        'technical_indicators': ...,  # RSI, MACD, SMA50/200, BB bands
        'performance':          ...,  # period_return, volatility, max/min
        'analyst_info':         ...,  # recommendation, target prices
    }

CLI Usage β€” Run Directly for Quick Testing

# Print console summary
python helper-script.py --symbol TSLA --period 6mo

# Generate interactive HTML report
python helper-script.py --symbol AAPL --period 1y --report

# Save raw analysis as JSON
python helper-script.py --symbol MSFT --period 2y --output msft.json

# Print rendered email summary
python helper-script.py --symbol NVDA --period 1y --email

7. πŸ”§ Installing the Skill in Claude Code

πŸ”§ Installing the Skill in Claude Code

Claude Code is the CLI that runs Claude directly from your terminal with native Skills support.

πŸ“š GitHub Repository

Step 1 β€” Prerequisite: Claude Code must already be running

npm install -g @anthropic-ai/claude-code
claude --version   # verify installation

Step 2 β€” Create the working directory structure

mkdir -p claude-workspace/.claude/skills
cd claude-workspace

Step 3 β€” Copy the stock-analyzer skill into the skills directory

cp -r /path/to/stock-analyzer .claude/skills/

Verify the structure looks exactly like this:

claude-workspace/
└── .claude/
    └── skills/
        └── stock-analyzer/
            β”œβ”€β”€ SKILL.md
            β”œβ”€β”€ helper-script.py
            β”œβ”€β”€ templates/
            └── resources/

πŸ’‘ The .claude folder is Claude Code’s project config home β€” like .git for Git. Every skill package goes directly inside .claude/skills/.

Step 4 β€” Install Python dependencies

pip install yfinance pandas numpy plotly

8. πŸš€ Run It β€” Analyze TESLA Stock

Open your terminal from claude-workspace/ and start Claude Code. Use /skills to verify your skill is discovered, then ask your first query.

Launch Claude Code

~/claude-workspace $ claude
βœ“ Claude Code started Β· project: claude-workspace
βœ“ Skills directory found: .claude/skills/

Check Available Skills

claude> /skills

πŸ“¦ Available Skills β€” 1 found in .claude/skills/

  01  stock-analyzer   Comprehensive stock market analysis using Yahoo Finance data

Type a query to auto-trigger a skill Β· or use /skill <name> to invoke directly

Ask Your First Query

claude> Analyze TESLA stock using stock-analyzer

⚑ Skill matched: stock-analyzer  (triggered by: "analyze TESLA stock")
πŸ“₯ Loading SKILL.md instructions...
🐍 Running: helper-script.py --symbol TSLA --period 1y --report
   Fetching data for TSLA...
   Computing RSI, MACD, Bollinger Bands, SMA 20/50/200...
   Building interactive Plotly charts...
   Rendering report from templates/report.html...

βœ… Analysis Complete β€” TSLA (Tesla Inc.) Β· 1 Year
   Current Price      $248.50
   1-Year Return      +38.2%
   RSI (14)           62.4 β€” Neutral
   MACD Signal        Bullish crossover
   SMA 50 / 200       $231.40 / $218.60
   P/E Ratio          78.5x
   Analyst            BUY (32 analysts)
   Report saved       TSLA_report_20260318.html

πŸ”§ Sample report output

Quick Reference

Command What it does
/skills List all skills in .claude/skills/
/skill stock-analyzer Directly invoke the skill by name
Natural language query Claude auto-matches to the right skill
.claude/skills/ Drop any skill folder here to activate it

βœ… Summary

πŸš€ That’s it! Claude read SKILL.md, followed its instructions, ran helper-script.py, and generated a full interactive HTML report β€” all from one natural-language query. This is the power of Claude Skills.

  • SKILL.md is the brain β€” write a clear description and Claude knows exactly when to trigger
  • helper-script.py is the engine β€” pre-tested functions, no code generation at runtime
  • templates/ are the blueprints β€” consistent professional output every time
  • resources/ are the reference β€” Claude looks up facts instead of guessing
  • .claude/skills/ is where Claude Code discovers your packages

πŸ“– Previous: Claude Skills β€” Introduction & Overview

πŸ›  Hands-On: Build a Stock Analyzer Expert with Claude Skills Β· aicodegeek.com