new standard for agents configuration and skills

Use multiple AI agents? Cursor, Claude code, Codex and more?

Writing skills files too I'll bet. Probably rules files as well.
Guess what. Now your skills are in multiple different folders such as .claude/skills or .cursor/skills etc. Same for commands, rules etc.

There's a new emerging standard for consolidating and sharing configuration and capabilities. It's a new folder .agents which would include skils subfolder and more.

Yeah great.

But sadly currently Claude Code doesn't load skills from .agents. It only reads from .claude folder.
So in Claude /mynew-skill won't resolve as a slash command, from .agents and the Skill tool can't invoke it.

Claude said if you want mynew-skill usable as /mynew-skill, the options are:

1. Symlink it: ln -s ../../.agents/skills/mynew-skill .claude/skills/mynew-skill — one source of truth, both tools see it.
2. Move it to .claude/skills/ and symlink the other direction for Codex.
3. Leave it and just tell me "follow .agents/skills/mynew-skill/SKILL.md" — I'll read and follow it, you just lose the slash command and the auto-trigger from the description.

Too bad. I want this to be vendor agnostic. Not maintain in multiple different places. It's a bit of. a mess currently.

I've been experimenting with pointing CLAUDE.md to AGENTS.md and put content in AGENTS.md which OpenAI, Anthropic and Cursor now all use.


Comments

Popular posts from this blog

angular js protractor e2e cheatsheet

angularjs ui-router query string parameter support

Notes on interview with Kent Beck about AI and why it's like a "genie"