Claude Code Trading Skills: How to Install and Use Them (2026)

What Are Claude Code Skills?

Claude Code skills are instruction files (SKILL.md) that tell Claude exactly how to handle a specific type of task. Think of them as saved recipes. Instead of writing a long prompt every time you want Claude to analyze a stock or place a trade, you install a skill once and it activates automatically when the task matches. Skills can call MCP tools, write code, read files, and execute multi-step workflows.

Why Skills Matter for Trading

Without a skill, you type something like: “Fetch SPY daily data from Alpaca, calculate RSI and MACD, check if RSI is below 30 and MACD just crossed bullish, and if so place a paper trade for 10 shares with a 2% stop loss.” You do this every time.

With a trading skill installed, you type: “Analyze SPY and trade if the setup is there.” The skill already knows your preferred indicators, risk rules, position sizing, and which broker to use. It runs the full workflow automatically.

The difference is not just convenience. Skills enforce consistency. A skill does not forget your stop-loss rule on the fifth trade because it got tired. It does not skip the volume check because the setup “looks good enough.” It follows the same process every single time.

Best Trading Skills Available in 2026

The Claude Code skills ecosystem for trading has grown fast since early 2026. Here are the skills worth knowing about.

VibeTrader

VibeTrader is the most popular trading skill right now. It connects Claude Code to Alpaca’s API and lets you build, backtest, and deploy strategies from natural language. You can go from “I want a momentum strategy on tech stocks” to a running paper trading bot in a single session.

What it covers:

Capability	Details
Strategy building	Describe a strategy in plain English, get working Python code
Backtesting	Run strategies against historical data with PnL reports
Paper trading	Execute on Alpaca paper accounts
Live trading	Switch to real execution (use with caution)
Portfolio monitoring	Track positions, PnL, and risk metrics
Options analysis	View chains, calculate Greeks, evaluate spreads

Install:

cd ~/.claude/skills
git clone https://github.com/vibetrader/vibetrader-skill.git vibetrader

Or create the skill file manually:

mkdir -p ~/.claude/skills/vibetrader

Then add a SKILL.md file in that directory following the project’s README instructions.

VibeTrader works best when paired with the Alpaca MCP server. The MCP server handles the API connection while the skill handles the trading logic.

claude-trading-skills (by tradermonty)

This is an open-source collection of five focused skills rather than one big skill. Each handles a specific part of the trading workflow:

Skill	What It Does
Market Analysis	Screens stocks by sector, volume, price action
Technical Charting	Calculates indicators (RSI, MACD, Bollinger, etc.) and identifies patterns
Economic Calendar	Pulls upcoming economic events and earnings dates
Screener	Filters stocks by customizable criteria
Strategy Development	Helps build and backtest rule-based strategies

Install:

cd ~/.claude/skills
git clone https://github.com/tradermonty/claude-trading-skills.git trading

These skills are lighter weight than VibeTrader. They are a good fit if you want individual tools you can mix and match rather than an all-in-one solution.

Autonomous Trading Skill

This skill focuses on running trading agents that operate without constant input. You define the strategy, risk parameters, and schedule, then the agent runs autonomously. It logs decisions and sends alerts when it takes action.

The key difference from VibeTrader is the emphasis on long-running autonomous operation. VibeTrader is interactive: you prompt it and it responds. The autonomous skill is designed to run in the background using Claude Code routines.

This skill pairs well with a VPS setup. See our guide on running trading bots on a Mac Mini for always-on hardware ideas.

How to Install Any Claude Code Skill

The process is the same for every skill:

Step 1: Create the Skills Directory

mkdir -p ~/.claude/skills

Step 2: Add the Skill

Either clone the repo:

cd ~/.claude/skills
git clone https://github.com/some-org/some-skill.git skill-name

Or create the SKILL.md file manually:

mkdir -p ~/.claude/skills/my-trading-skill
nano ~/.claude/skills/my-trading-skill/SKILL.md

Step 3: Verify in Claude Code

Start a new Claude Code session and ask:

What skills do you have available?

Claude should list the installed skill. Then test it:

Use the [skill-name] skill to analyze AAPL.

Building Your Own Trading Skill

Off-the-shelf skills work for general use cases, but serious traders want custom logic. Here is how to build a skill tailored to your strategy.

SKILL.md Structure

A trading skill file has three parts: description, tools, and instructions. Here is a template for a simple swing trading skill:

# Swing Trading Analyst

## Description
Analyze stocks for swing trading setups on the daily timeframe.
Focus on pullbacks to support in uptrending stocks.

## Tools Required
- Alpaca MCP server (for market data and execution)
- Python (for indicator calculations)

## Instructions

### When the user asks to analyze a stock:
1. Fetch 200 daily bars using Alpaca MCP
2. Calculate these indicators:
   - 20-day and 50-day SMA
   - 14-period RSI
   - Average True Range (ATR) for position sizing
3. Check for a swing setup:
   - Price is above the 50-day SMA (uptrend confirmed)
   - Price has pulled back to within 1 ATR of the 20-day SMA
   - RSI is between 40-55 (pullback, not oversold)
4. If setup is present, calculate:
   - Entry: current price
   - Stop loss: 1.5x ATR below entry
   - Target: 3x ATR above entry (2:1 reward/risk)
   - Position size: risk 1% of account on the trade
5. Present the analysis with a clear BUY or WAIT recommendation
6. If user confirms, place the order through Alpaca MCP

### Risk Rules (never override these):
- Maximum 5 open positions at any time
- No single position larger than 20% of account
- No trading in the first 15 minutes after market open
- No new positions on FOMC announcement days

Testing Your Skill

Start a Claude Code session and try it:

Analyze MSFT for a swing trade setup.

If the skill is working, Claude will follow the exact process you defined: fetch data, calculate indicators, check the setup criteria, and give you a recommendation with specific numbers.

Refine the instructions based on real usage. If Claude skips a step or makes wrong assumptions, update the SKILL.md to be more explicit about that step.

Combining Skills with MCP Servers

Skills and MCP servers serve different purposes. MCP servers provide the data connection (market prices, account info, order execution). Skills provide the trading logic (when to buy, how much, where to set stops).

A strong AI trading setup in 2026 looks like this:

Layer	Tool	Role
Data and execution	Alpaca MCP V2	Stocks, options, crypto
Data and execution	Bybit MCP	Crypto futures
Chart analysis	TradingView MCP	Visual chart reading
Trading logic	VibeTrader or custom skill	Strategy, signals, risk rules
Research	Alpha Vantage MCP	Fundamentals, earnings data

You configure the MCP servers in ~/.claude/.mcp.json and the skills in ~/.claude/skills/. Claude Code uses both layers together. The skill tells it what to do, and the MCP server gives it the tools to do it.

For MCP server setup details, see our MCP servers guide.

What Changed for Skills in June 2026

Skills are still in beta, and the June Claude Code release tweaked a few things that affect trading skills specifically:

Change	What It Means for a Trading Skill
Literal `$` escape syntax	You can now put a literal `$` before a digit in a command body by escaping it. Handy when a skill prints dollar amounts like `$5` in a log line without Claude treating it as a variable
Skills run beyond Claude Code	Skills are showing up in more Anthropic surfaces, so a SKILL.md you write for trading can travel to other clients instead of being locked to one terminal
Lazy-loaded MCP tool catalogs	MCP tool search now loads large tool catalogs on demand. A skill that leans on a 60-tool broker server no longer pays the full catalog cost up front

None of these change how you write a trading skill, but the $ escape is worth knowing if your skill logs prices or PnL. Before the fix, a log template like Filled at $5,420 could get mangled when the digit right after the dollar sign looked like a positional argument. Escaping it keeps the output clean.

The lazy-loading change matters more as your setup grows. If you run Alpaca MCP V2 (61 tools) plus a chart server plus a data server, the skill used to feel the weight of every tool definition at once. Now Claude pulls the tool schemas it actually needs for the task in front of it, which keeps a focused skill — “analyze AAPL and place one trade” — fast even with a large server attached.

Common Mistakes with Trading Skills

Being too vague in instructions. “Analyze the stock and decide if it’s a good buy” gives Claude too much room to make things up. Specify exactly which indicators, what thresholds, and what the output format should look like.

Not setting hard risk limits. Every trading skill should have a “never override these” section with position sizing limits, maximum open positions, and daily loss limits. Without hard constraints, a creative AI might rationalize oversized positions.

Skipping the paper trading phase. Install the skill, connect it to a paper trading account, and run it for at least two weeks before going live. Watch for edge cases: how does it handle gaps? What about halted stocks? Does it behave correctly around earnings dates?

Installing too many overlapping skills. If two skills both try to analyze stocks with different criteria, Claude may get conflicting instructions. Keep your installed skills focused on different tasks or use a single comprehensive skill.