AI agent skills are becoming a useful packaging format. A good skill is not just a prompt. It is a small folder that usually contains: a SKILL.md instruction file; scripts; examples; docs; repeatable commands; tool-specific setup notes. That makes skills useful for real agent workflows. But it also creates a distribution problem. Where should a skill live? How should a user install it? How does the same skill reach OpenClaw, Hermes, Claude Code, Codex, Cursor, OpenCode, or other agent runtimes? One interesting answer is gh skill , a GitHub CLI extension by Nicholas Spencer. Repo: https://github.com/nicholasspencer/gh-skill Enter fullscreen mode Exit fullscreen mode The idea is simple: Use GitHub Gists as a lightweight registry for AI agent skills. A Gist already supports multiple files, version history, forks, stars, and GitHub API access. gh skill adds a convention on top so a Gist can behave like a skill package.…