Skip to content
CodeLifter

Agent skills directory

A skill is a folder of instructions, and sometimes scripts, that an assistant loads when a task calls for it. The format is an open standard published at agentskills.io. Each listing below says who publishes it, which licence covers it and whether it ships code for the assistant to run.

A folded blank sheet beside an open tool roll of hand tools, one with an orange-red handle

Short answer

A skill is a folder with a SKILL.md file whose front matter gives a name and a description. The assistant reads only those two fields at startup and loads the full file when a task matches.

Install one by copying its folder into a skills directory. Claude Code reads ~/.claude/skills/ for personal skills and .claude/skills/ in a project. Codex reads ~/.agents/skills/ and .agents/skills/ from the working directory up to the repository root. GitHub Copilot reads project skills from .github/skills/, .claude/skills/ or .agents/skills/, and personal skills from ~/.copilot/skills/ or ~/.agents/skills/. The name must be lowercase letters, numbers and hyphens, and must match the folder name. Read any bundled scripts first: a skill can tell the assistant to run them.

assisted-by-trailer/SKILL.md Markdown · 14 lines Reviewed
---
name: assisted-by-trailer
description: Use when committing code an AI assistant helped write. Adds an Assisted-by trailer and a pull request note.
---
# Assisted-by trailer
When you commit changes that include generated code:
1. End the message with one trailer line:
Assisted-by: <tool name> (<what a human checked>)
2. In the pull request, list which files were generated
and how each was reviewed.
3. Never remove an existing Co-authored-by line.
An example skill written by the editors, not a published package. The front matter (name and description) is all the assistant sees before it loads the rest.

Skills 10

  1. skill-creator

    Meta

    Creates new skills and edits existing ones, then runs evals to measure how well a skill works and how reliably its description triggers. Ships Python scripts for benchmarking and reports.

    Runs scripts Apache-2.0 anthropics/skills

  2. mcp-builder

    Dev tools

    A guide to building MCP servers in Python (FastMCP) or TypeScript (the MCP SDK). Includes an evaluation script for testing a finished server.

    Runs scripts Apache-2.0 anthropics/skills

  3. webapp-testing

    Dev tools

    Tests local web apps with Playwright: checks frontend behaviour, captures screenshots and reads browser logs. A helper script starts and stops your dev server.

    Runs scripts Apache-2.0 anthropics/skills

  4. frontend-design

    Design

    Guidance on aesthetic direction and typography for new or reworked interfaces, so they do not read as templated defaults. The folder holds no code.

    Instructions only Apache-2.0 anthropics/skills

  5. web-artifacts-builder

    Dev tools

    Builds multi-component HTML artifacts with React, Tailwind CSS and shadcn/ui, using shell scripts to set up the project and bundle it into one HTML file.

    Runs scripts Apache-2.0 anthropics/skills

  6. docx

    Documents

    Creates, reads and edits Word documents and templates, including tracked changes and comments. Python scripts convert, merge and validate files.

    Runs scripts Proprietary anthropics/skills

  7. pdf

    Documents

    Extracts text and tables, merges and splits files, fills forms and runs OCR on scans. Form filling follows bundled Python scripts.

    Runs scripts Proprietary anthropics/skills

  8. xlsx

    Documents

    Opens, edits and builds spreadsheets (.xlsx, .csv, .tsv) with formulas and formatting. A bundled script recalculates formulas after edits.

    Runs scripts Proprietary anthropics/skills

  9. pptx

    Documents

    Creates, reads and edits slide decks. Bundled scripts add slides, remove unused files and render thumbnail grids of every slide.

    Runs scripts Proprietary anthropics/skills

  10. brand-guidelines

    Design

    Applies Anthropic’s own brand colours and type (Poppins headings, Lora body) to what the assistant makes. Most useful as a pattern for writing a skill for your own brand.

    Instructions only Apache-2.0 anthropics/skills

Checked on 2026-10-15 against each skill’s SKILL.md and LICENSE.txt in the anthropics/skills repository. Licences differ by folder: six of these ten are Apache-2.0, while docx, pdf, pptx and xlsx carry a proprietary licence (the repository calls them source-available, not open source). “Runs scripts” means the folder ships code that its instructions tell the assistant to run.

Skill, MCP server or prompt?

All three shape what an assistant does. They differ in what they can reach.

Prompt
Text you paste once. Reaches nothing on its own.
Skill
Instructions and scripts loaded when relevant. Runs with the assistant’s existing tools.
MCP server
A program that adds new tools: files, APIs, browsers. Reaches whatever you grant it.