skilltags

Generate slash commands that help agents use the right installed skills.

Tired of hunting for the right skill name every time you want the agent to actually use the skills you've installed?

skilltags generates Cursor command files like /st-frontend, /st-backend, and /st-design. Each command file contains a curated list of matching skills plus instructions telling the agent to review that list before it starts work.

Built for the skills.sh ecosystem.

Quick Start

1. Install

npm install skilltags -g

Global install is the best default if you want the skilltags command in your shell.

If you want project-only setup instead, local install works too:

npm install skilltags

Global install creates or updates files in ~/.cursor/. Local install creates or updates files in .cursor/ inside the current repo.

On a fresh install, an interactive setup wizard walks you through category selection. Global install can also enable auto-sync. If you've already configured that scope, postinstall silently re-syncs it.

If the wizard didn't run, configure manually:

skilltags update
npx skilltags update --local

2. Pick a scope and categories

Choose the scope you want:

• Global: ~/.cursor/skilltags.json and ~/.cursor/commands/st-*.md
• Project: .cursor/skilltags.json and .cursor/commands/st-*.md

Then pick the categories that fit your workflow: frontend, backend, design, and so on. skilltags scans known skill directories, matches skills to those categories, and writes one Cursor command file per category.

3. Use in prompts

Add a category command to the end of any Cursor prompt:

I want to make my website components look more modern and responsive.
/st-frontend

The agent reads the generated st-frontend.md command file, reviews the listed skills, and opens the relevant ones before starting work.

Tip: New to skills? Browse and install from skills.sh:

npx skills find            # search the skills directory
npx skills add owner/repo  # install a skill package

The problem

You install great skills, but when you prompt the agent, it just starts coding. It skips the skills entirely unless you explicitly tag each one by name. You end up hunting for exact skill names every single time.

The fix

Categories

Add a category command to the end of your prompt using /:

Don't see a category you need? Suggest one. We're always expanding the list.

How It Works

Install -> Choose Global or Project Scope -> Select Categories -> Scan Known Skill Paths -> Match Skills -> Write skilltags.json + st-*.md Files

What's inside each command file

Each generated st-{category}.md file contains two parts:

1. An instructional header that tells the agent to review the listed skill names and descriptions, identify which ones are relevant to the user's request, and open those skills before starting work.
2. A curated skill list with every matched skill for that category, including its name, file path, and description.

Important: The instructional header is what makes this work. It tells the agent to evaluate the listed skills against the request first, instead of jumping straight into implementation.

Commands

skilltags                      Sync category files from current config
skilltags sync                 Same as above (explicit)
skilltags update               Add or remove skill categories
skilltags update <category>    Edit skills within a specific category

Full reference: docs/usage.md

Contributing

Suggest a category. Think a new skill category would be useful? Open an issue with the category name and what types of skills it should cover. The more specific the better. Including "what keywords should match this category?" helps us get it right.

Improve keyword matching. Category keyword mappings live in lib/categories.js. If a skill isn't landing in the right category, PRs to improve the keyword lists are welcome.

_{MIT License}