Agent Skills Overview¶

Agent Skills is a simple, powerful way to specialize agents using markdown files. This guide covers everything you need to know to use skills in your Swarms agents.

What are Agent Skills?¶

Agent Skills are modular capabilities defined in SKILL.md files that guide agent behavior. They enable you to:

Specialize agents without modifying code
Share expertise across your team through skill libraries
Maintain consistency in how agents perform tasks
Rapidly iterate by editing markdown instead of rebuilding prompts

Quick Example¶

from swarms import Agent

agent = Agent(
    agent_name="Financial Analyst",
    model_name="gpt-4o",
    skills_dir="./example_skills",  # ← Just add this parameter!
    max_loops=1
)

# Agent automatically follows financial-analysis skill instructions
response = agent.run("Perform a DCF valuation for Tesla")

SKILL.md Format¶

Skills use a simple structure:

---
name: my-skill
description: What this skill does
---

# Skill Instructions

Detailed guidance for the agent...

## Guidelines
- Key principles
- Best practices

## Examples
- Example use case 1
- Example use case 2

Directory Structure¶

skills/
├── financial-analysis/
│   └── SKILL.md
├── code-review/
│   └── SKILL.md
└── customer-support/
    └── SKILL.md

Core Concepts¶

1. Skill Loading¶

When you specify skills_dir, Swarms: 1. Scans for subdirectories containing SKILL.md 2. Parses YAML frontmatter (name, description) 3. Loads full markdown content 4. Injects into agent's system prompt

2. Skill Activation¶

Skills are always active once loaded. The agent sees all skill instructions and applies them when relevant to the task.

3. Multiple Skills¶

Agents can have multiple skills loaded simultaneously. They'll intelligently apply the right guidance based on the task.

Example Skills Included¶

Swarms includes 3 production-ready skills:

Financial Analysis¶

example_skills/financial-analysis/SKILL.md

- DCF valuation methodology - Financial ratio analysis - Sensitivity analysis frameworks - Investment recommendations

Code Review¶

example_skills/code-review/SKILL.md

- Security vulnerability detection - Performance optimization checks - Best practices enforcement - Maintainability assessment

Data Visualization¶

example_skills/data-visualization/SKILL.md

- Chart type selection - Design principles - Color best practices - Storytelling frameworks

Common Use Cases¶

1. Domain Expertise¶

Add specialized knowledge:

# Financial analysis agent
agent = Agent(
    agent_name="Finance Expert",
    skills_dir="./skills/finance"
)

2. Process Enforcement¶

Ensure consistent methodologies:

# Code review with company standards
agent = Agent(
    agent_name="Code Reviewer",
    skills_dir="./skills/code-standards"
)

3. Communication Styles¶

Define tone and formatting:

# Customer support with brand voice
agent = Agent(
    agent_name="Support Agent",
    skills_dir="./skills/support"
)

Next Steps¶

Explore the example guides:

Basic Skills Usage - Start here for your first skill
Creating Custom Skills - Build your own skills
Multiple Skills - Use multiple skills together

Key Benefits¶

✅ Simple: Just markdown files, no code needed
✅ Portable: Standard format works across platforms
✅ Version Control: Skills are just files - track with git
✅ Reusable: Share skills across your organization
✅ Claude Code Compatible: Works with Anthropic's ecosystem

Compatibility¶

Swarms Agent Skills implementation follows Anthropic's Agent Skills standard, ensuring compatibility with Claude Code and other tools that use the same format.

What This Means¶

Skills created for Swarms work with Claude Code
Skills from Claude Code work with Swarms
Same SKILL.md format across platforms
Part of an emerging ecosystem standard