Agent Skills
January 30, 2026
Agent Skills are portable packages of instructions, scripts, and resources that give AI agents new capabilities. Think of them as plugins for agents — reusable, shareable, and discoverable.
The Core Concept
Skills solve a fundamental problem: agents are increasingly capable, but often lack the context they need to do real work reliably. Skills provide procedural knowledge — step-by-step instructions for specific tasks — that agents can load on demand.
/skills/
/next-js-docs/
SKILL.md # Instructions for the agent
examples/ # Code snippets
api-reference/ # Compressed documentation
Unlike AGENTS.md (which provides passive context), skills are active: agents discover them, decide when to invoke them, and load the relevant content.
What Skills Enable
Skills extend agent capabilities across several dimensions:
- Domain expertise: Package specialized knowledge — legal review processes, data analysis pipelines, framework-specific patterns
- New capabilities: Give agents abilities like creating presentations, building MCP servers, analyzing datasets
- Repeatable workflows: Turn multi-step tasks into consistent, auditable processes
- Interoperability: Reuse the same skill across different agent products
Anatomy of a Skill
A skill is a folder containing a SKILL.md file and supporting resources:
# SKILL.md
## Overview
This skill helps agents build Next.js applications using the App Router.
## When to use
- Creating new Next.js projects
- Migrating from Pages Router to App Router
- Implementing server components
## Instructions
1. Check the Next.js version in package.json
2. Use Server Components by default
3. Colocate data fetching with components
...
Skills can include:
- Documentation and examples
- Scripts the agent can execute
- Templates and boilerplate
- Test cases for validation
The Tradeoff: Skills vs AGENTS.md
Vercel's research revealed a crucial finding: in evals, agents invoked skills only 44% of the time — even when those skills would have helped.
This doesn't mean skills are useless — it means they're complementary. As one developer explained:
Tools are just functions under the hood. Skills are markdown files with instructions instead. Both get interpreted into prompts. The agent harness loads them at different times depending on which one you choose.
When to Use Skills
Skills are the right choice when:
- Scale: The knowledge base is too large for context
- Specialization: Knowledge is only relevant for rare tasks
- Capabilities: The agent needs to do something, not just know something
- Actions: Skills can trigger tools, scripts, and external systems
MCP for broad integration, Skills for task-specific boosts. The "dead" claim seems exaggerated; both have roles.
Supported Agents
Skills are supported by leading AI development tools:
- Claude Code — Native skills support via Anthropic
- Cursor — Skills integration
- Amp — Built with skills compatibility
- Cline — Skills-aware
Creating Skills
The Agent Skills format was originally developed by Anthropic, released as an open standard, and has been adopted by a growing number of agent products.
Basic structure:
/my-skill/
SKILL.md # Required: instructions
/examples/ # Optional: code samples
/templates/ # Optional: scaffolding
/docs/ # Optional: reference material
The SKILL.md file should include:
- Overview: What the skill does
- When to use: Trigger conditions
- Instructions: Step-by-step guidance
- Examples: Concrete demonstrations
Skills vs MCP
Skills and MCP serve different purposes:
- Focus — Skills: instructions & knowledge. MCP: tools & data access.
- Type — Skills: Markdown + files. MCP: protocol + servers.
- Activation — Skills: agent decides to load. MCP: always available.
- Use case — Skills: "How to do X". MCP: "Connect to Y".
Many workflows combine both: MCP provides the connection layer, skills provide the know-how.
Sources
- Agent Skills — Official specification
- GitHub: agentskills/agentskills — Open standard
- AGENTS.md vs Skills — Vercel's research
See also: AGENTS.md · MCP Servers · Context Loading