Claude Skills
TL;DR Highlight
Anthropic introduced 'Skills' in Claude — a folder bundling instructions, scripts, and resources to give the LLM new capabilities.
Who Should Read
Developers and power users who want to extend Claude's capabilities beyond default behavior with reusable, composable skill packages.
Core Mechanics
- Claude Skills are self-contained packages: a markdown instruction file + optional scripts + optional resource files in one folder
- Skills can be loaded per-conversation, enabling modular capability extension without system prompt bloat
- Enables sharing and reusing capabilities across teams without re-prompting from scratch
- Works within the existing Claude API / Claude.ai interface
- A step toward agent-like extensibility: skills can encapsulate tools, workflows, and domain knowledge
Evidence
- Announced by Anthropic with initial documentation and examples
- Community testing shows skills load reliably and follow instructions in the markdown file
- Early adopters report significant workflow acceleration for repetitive LLM tasks
How to Apply
- Create a skill folder with a SKILL.md describing the capability and any helper scripts, then load it at the start of relevant sessions.
- Package domain-specific knowledge (e.g., your codebase conventions, API docs) as a skill to avoid repeating context each session.
- Share skill folders across your team as reusable building blocks for common LLM workflows.
Terminology
Claude SkillsAnthropic's system for packaging reusable LLM capabilities as a folder containing markdown instructions and supporting scripts/resources.
System PromptInstructions given to an LLM at the start of a conversation that define its behavior, persona, and constraints.
Modular CapabilityBreaking LLM capabilities into independent, composable units that can be mixed and matched per use case.