Hey, thanks for the report. I’d like to reproduce this on my side. Can you share a couple details?
Where did the created file end up, in .cursor/skills/ or .cursor/agents/ or in home ~/.cursor/...? A screenshot of the path or the file explorer would help a lot.
When the chat opened after clicking New in the Subagents section, did the prompt start with /create-subagent ...? Or did it show /create-skill? A screenshot of the composer window before you sent it is fine too.
What exact prompt did you type after the prefix?
From the description, this sounds like an edge case in how the model interpreted it, since the button should route to the right command, but without a repro it’s hard to say for sure. Let me know and we can dig deeper.
Thanks for the details, that’s enough. Confirming: the prefix was correct (/create-subagent), but the model still went into creating a skill. This looks like an issue on our side, I’ve already reported it internally. I can’t share an ETA for a fix yet.
Current workaround is to create the subagent manually:
Create a .cursor/agents/ folder in your project root (or ~/.cursor/agents/ globally).
Add a file called <name>.md with frontmatter like this:
---
name: my-subagent
description: When to call this subagent
---
Put the subagent system prompt here.
If a file was already created under .cursor/skills/<name>/SKILL.md, you can move it to .cursor/agents/<name>.md and keep only the frontmatter plus the prompt body (without the skill-specific structure).