Organizing Cursor skills for better agent context

Hey everyone,

If you’ve ever struggled with getting the Cursor agent to identify and actually use your installed skills when they are relevant, I built a lightweight, open-source CLI tool to solve this: skill-tags.

It scans your local skills and generates a consolidated @skill-tags.md file. By simply referencing this file in your chat, you provide the agent with a complete directory of your available tools, allowing it to autonomously assess and apply the right ones for your request.

The Highlight: Automatic Categorization CLI

If you have a massive library of skills and want to keep your context window focused without attaching the global index, you can run the interactive wizard:

skill-tags --categories

This automatically groups your skills into focused command files (like @skills-frontend.md, @skills-testing.md, or @skills-ai-agents.md) using strict keyword matching. This lets you attach only the domain-specific skills relevant to your current task so the agent knows exactly what to pull from.

Quick Start:

npm install skill-tags -g

You can check out the source code and documentation here:

Would love to hear any feedback or ideas from the community!