content-variants

I was digging through how skills handle content generation, and I noticed your content-variants approach tackles a pretty common problem—managing LLM output consistency across different contexts. The 72 score suggests you're in that tricky middle ground where the core idea is solid but there's room to tighten up the implementation details.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 72/100, which puts you in C territory—adequate work with real gaps. This is based on Anthropic's best practices for skill architecture. Your strongest area is Spec Compliance (12/15, solid YAML and naming), but Progressive Disclosure (18/30) is dragging you down. You've got good bones here, but the skill needs external reference files and clearer workflow guidance to really shine.

What's Working Well

• Solid spec compliance – Your frontmatter is clean, YAML is valid, and naming conventions follow the hyphen-case standard perfectly
• Clear problem domain – The framework structure (Variant Inventory → Versioning → Testing → Approval → Performance) makes logical sense
• Consistent terminology – You use 'variant', 'asset', and 'approval' consistently throughout, which helps readers follow along
• Grep-friendly structure – Clean, scannable format that works well in search contexts

The Big One: Missing Reference Files

This is what's really holding you back. Your Templates section lists three key templates as bullet points:

• Variant matrix (channel × persona × lifecycle stage)
• Approval checklist
• Performance tracker

But these don't actually exist as separate files. This kills your PDA score because you're trying to cram guidance into inline descriptions instead of building a layered structure. The fix: Create references/ directory with three markdown files:

• references/variant-matrix-template.md (with concrete example: channels, personas, actual data)
• references/approval-checklist.md (actual checklist items, not just concepts)
• references/performance-tracker.md (sample tracking table with real metrics)

This alone would bump you +5 points into the mid-70s by improving PDA and Utility.

Other Things Worth Fixing

1. Vague trigger phrases – Your description says "planning, approving, versioning" but these don't match how people actually search. Add specific triggers like "create variant matrix", "track variant performance", "audit creative inventory" in your frontmatter description.

2. Framework lacks actionable steps – "Catalog base asset, variant name, audience..." reads like a checklist of fields rather than steps. Break it into actual actions: "Create spreadsheet → Document base asset ID → Assign unique variant name → Tag audience segment" etc.

3. No validation section – You mention a "Performance tracker" but no guidance on how to validate variants work or check for completeness. Add a validation section covering: completeness checks, CDN verification, compliance audit, expiration flagging.

4. Inconsistent voice – Mixing "catalog base asset" with "note experiment IDs" creates friction. Pick imperative throughout ("Catalog...", "Note...") or noun form throughout.

Quick Wins

• Add three reference files with actual templates and examples (+5 points)
• Rewrite description with specific trigger phrases like "create variant matrix" (+3 points)
• Convert framework bullet points into numbered action steps (+2 points)
• Add validation/feedback loops section (+2 points)

These moves would realistically push you into B territory (80+) without major restructuring.

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.