Browse reusable Agent Skills for Codex, Claude, document work, development, research, and repeatable AI workflows. Compare compatibility, setup, licensing, and official sources.
How to choose
Every entry links to the maintainer's source and records compatibility, license, runtime, and review date. Confirm permissions and current release notes before using it in production.
An Agent Skill is a reusable folder of instructions and supporting resources that teaches an AI agent how to complete a specialized workflow. A good skill does more than hold a long prompt: it defines when the capability should activate, what inputs it needs, which tools or scripts it may use, how to check the output, and what boundaries it must respect.
Skills are useful when a workflow repeats. Examples include creating a branded document, checking a pull request, researching official product documentation, preparing a recurring report, or applying a team-specific review process. They make agent behavior more consistent without forcing every instruction into the main conversation.
Start with compatibility. Confirm that the skill format works with your agent host, such as Codex or Claude Code. Then review the complete source, not only the skill description. Look for scripts, network calls, required credentials, file access, and referenced assets. Check the license for every folder you intend to reuse because collections can contain a mix of open-source and source-available material.
Prefer skills with a narrow trigger, explicit inputs, verifiable outputs, and a recent review date. A skill that claims to handle every task usually provides less predictable behavior than one designed around a concrete workflow. For team use, assign an owner and test the skill against a small set of representative tasks before publishing it broadly.
A prompt is usually the smallest building block: it shapes one model response. A skill packages a repeatable method. An MCP server gives the agent controlled access to an external tool or data source. Many reliable workflows combine all three, but you should begin with the smallest layer that solves the task. Browse the complete AI resource library, connect external systems through the MCP server directory, or compare orchestration options in the agent framework directory.
Curated directory
A public collection of Agent Skills covering documents, creative work, development, communication, and skill templates.
A maintained Codex skill that routes OpenAI product questions to current official documentation and reference material.
An official system skill for designing, structuring, and validating reusable Codex skills.