Design Atlas
Type
agent skill collection
Verdict
Useful
Licence
Ship licence, open source permissiveopen-source-permissivefree; MIT (repo shadcn-labs/skills, about 24 stars, last push 2026-09-24). skills.sh counted 133 installs in total at review: tailwind-to-stylex 49, launch-shadcn-registry 33, mastra-file-agents 18, and 11 each for the three icon skills.
Agent channels
skill
Pricing
free
Reviewed

← Atlas · Topics: agent-skills, icons, components

Shadcn Labs Skills ​

Type
agent skill collection
Verdict
Useful
Licence
Ship licence, open source permissiveopen-source-permissivefree; MIT (repo shadcn-labs/skills, about 24 stars, last push 2026-09-24). skills.sh counted 133 installs in total at review: tailwind-to-stylex 49, launch-shadcn-registry 33, mastra-file-agents 18, and 11 each for the three icon skills.
Agent channels
skill
Pricing
free
Reviewed

What it is ​

This is the skills repo of Shadcn Labs, the group behind termcn, pdfcn and other shadcn-style registries. It holds six skills in four groups. launch-shadcn-registry checks a custom registry's registry.json, prepares pull requests to directories (the official shadcn list, registry.directory, awesome-shadcn-ui, free-for-dev, awesome-ai-devtools, shadcntemplates), and drafts launch posts for Reddit, X, Dev.to and Hacker News. The three icon skills make, audit and extend SVG icon sets against a fixed spec, using Python scripts that need only the standard library. tailwind-to-stylex moves Tailwind classes to StyleX. mastra-file-agents moves Mastra agents into one folder per agent. Only the first is about shadcn itself.

When to open it ​

  • When you have built a shadcn registry and want an agent to handle the tedious listing and announcement work.
  • When a product needs original icons that look like one family, or an existing set has drifted and needs an audit.
  • When you need to add icons to Lucide, Heroicons or Phosphor that match the originals.

Most useful ​

  • icon-set-generator: fixes stroke, grid, optical size and shared parts before drawing, then validates the batch and builds a preview page.
  • icon-set-audit: reports drift, divergent repeated parts, near-duplicates and size outliers in priority order.
  • icon-set-extend: infers the spec from existing files so new icons blend in.
  • launch-shadcn-registry: a profile schema plus a five-phase checklist from preflight checks to post-launch follow-up.
  • tailwind-to-stylex: resolves each utility to CSS before rewriting it as stylex.create styles.

Using it with agents ​

Install everything with npx skills add shadcn-labs/skills, or pick one with --skill <name>. The skills follow the Agent Skills format, so they work in Claude Code, Cursor, Codex and other supporting agents. Each has a SKILL.md, reference files and an evals folder, and the icon and launch skills also ship scripts. The launch skill needs network access, curl, git and a logged-in gh, and it says to open pull requests only after the user approves.

Watch out for ​

  • Install counts are very low, so these have seen little outside testing.
  • The launch skill opens PRs and drafts public posts. Keep a human check on anything it sends, and follow each directory's own submission rules.
  • The Python scripts are large (the audit script is about 47 KB). Read them before running them on a big repo.
  • The repo also holds internal helper skills under .agents/ that are not part of the published set.

Reusable ideas ​

  • Freeze an icon style spec (grid, stroke, gaps, recurring parts) before drawing anything, then validate against it.
  • Treat an icon library as a set of reused parts, so audits can find shapes that should match but don't.
  • Turn a launch checklist into a data profile plus one step per target, so the same facts feed every directory.
  • Make a migration skill resolve each utility to real CSS first, instead of mapping class names by pattern.

Shadcn Labs, startercn, Iconoir, shadcn/ui, Anthropic Skills