shadcn

I took a look at your shadcn skill and wanted to share some thoughts.

Links:

• GitHub
• Your skill in the SkillzWave marketplace

The TL;DR

You're at 87/100, solid B territory. This is graded against Anthropic's best practices for skills. Your strongest area is Ease of Use (23/25) — the metadata is tight, your triggers are well-chosen, and developers will find this skill easily. Weakest area is Spec Compliance (13/15), mostly because the description lacks trigger phrases.

What's Working Well

• Triggers are chef's kiss — You've got regex patterns like cn( catching actual code artifacts, plus intent triggers like "utility classes" and "component styling." That's the kind of specificity that gets skills activated when people actually need them.
• Reference structure is clean — SKILL.md stays under 220 lines while you ship 3 reference files + 5 templates. That's solid Progressive Disclosure — users get the quick start without drowning in options.
• Practical constraints — Your "Banned" section actually makes sense (no Radix UI reimplements, no copy-paste theming). Not just restrictions for restrictions' sake.
• Real examples throughout — The CVA pattern examples and accessibility patterns reference show you're solving genuine shadcn/ui pain points.

The Big One: Redundant Anti-Patterns Content

Here's what's eating into your PDA score: you've got anti-patterns documented both inline in SKILL.md (lines 142-150) and as a separate reference file. That's repetition that could be cleaner.

The fix: Remove the inline anti-patterns section from SKILL.md entirely. Replace it with a single pointer: "See references/anti-patterns.md for common mistakes to avoid." This reclaims tokens and keeps the main file focused on the workflow, not the gotchas.

Impact: +1 point easily, gets you closer to 88.

Other Things Worth Fixing

1. Voice inconsistency in mcp_first section — You write "Use natural language:" but the rest of SKILL.md is imperative. Make it "Natural language examples:" instead. Small, but it tightens the writing style (+0.5 points).

2. Description needs more trigger phrases — Right now you've got great triggers in the metadata section, but the description itself only hints at 1-2 of them. Weave in phrases like "component discovery" or "CSS-in-JS" to boost discoverability.

3. Success criteria lacks a verification loop — You've got a checklist, but no explicit "run → check → fix" pattern. Add: "Verification: Run build, check for TypeScript errors, confirm component renders correctly." This gives users a concrete way to validate they've done it right (+1 point).

4. Minor: XML tags penalty — You're using <critical> and similar XML tags in the metadata section. The spec doesn't need those — plain text triggers work just fine and save a few bytes.

Quick Wins

• Remove inline anti-patterns → reference the file instead (+1pt, cleaner PDA)
• Add trigger phrases to description → boosts discoverability
• Tighten voice → "Natural language examples" instead of "Use natural language" (+0.5pt)
• Add verification loop → gives users confidence they did it right (+1pt)

Those three moves get you to 90, solidly into A territory.

Checkout your skill here: SkillzWave.ai | SpillWave We have an agentic skill installer that install skills in 14+ coding agent platforms. Check out this guide on how to improve your agentic skills.