agentskills.codes

agentskills.codes is an open registry of installable Agent Skills for AI coding agents — Claude Code, Codex, GitHub Copilot, and others. Skills are reusable, portable capabilities an agent can be equipped with; the registry makes them browsable and installable with a straightforward command, the way package registries did for code libraries.

What does it offer?

The catalog spans the practical range of agent work: coding, research, design, security, documentation, and automation. Each skill page links back to its source and — notably — surfaces source-health information and capability scans before installation, so a developer can see what a skill actually does and where it comes from before granting it a place in their agent's toolkit. That pre-install transparency addresses the obvious worry with community-sourced agent instructions: installing a capability is an act of trust, and the registry puts evidence ahead of the install command.

Who is it for?

Developers and teams using AI coding agents day to day, who keep solving the same problem — teaching the agent a workflow — and want to reuse solutions rather than re-prompt them. Because the skills are portable across agents, teams mixing Claude Code, Codex, and Copilot can share one library instead of maintaining per-tool instructions. The registry is free and open source, which fits its stated goal: reusable agent workflows without vendor lock-in.

When would someone use it?

At the moment an agent needs a capability someone has probably already built — a security review checklist, a documentation pipeline, a research routine. Checking the registry first turns an afternoon of prompt engineering into a browse, a source check, and one install command.