hypothesis-library

I've been diving into property-based testing frameworks lately, and your hypothesis-library skill caught my attention—though the 50/100 score suggests there might be some gaps between the concept and execution that are worth digging into.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

TL;DR

You're at 50/100, solidly in F grade territory. This is based on Anthropic's skill evaluation best practices across five pillars. Your Spec Compliance is actually solid (11/15)—the frontmatter is valid and naming conventions are right—but Progressive Disclosure Architecture (12/30) and Utility (8/20) are where you're losing the most points. The core issue: you've got a great framework concept, but it's missing the teeth to be actually useful.

What's Working Well

• Consistent terminology: You use 'hypothesis', 'learnings', and 'experiment' consistently throughout—no confusing terminology shifts that would make users stumble.
• Solid metadata schema thinking: The idea of using ID, theme, persona, funnel stage, and metrics is the right foundation for structured experimentation tracking.
• Logical section flow: "When to Use" → "Framework" → "Templates" → "Tips" follows a reasonable progression that's easy to scan.

The Big One: Missing Reference Files Kills Progressive Disclosure

Your skill lists three templates—intake form, learning card, portfolio dashboard—but provides zero actual content. This is a critical gap. You're telling users "here are templates" without showing them what they look like, which means they either guess or bounce.

The fix: Create three reference files:

• references/intake-form.md – actual template with example fields
• references/learning-card.md – structured format showing context/result/recommendation
• references/portfolio-dashboard.md – example dashboard structure

This alone could add 7 points and make the skill actually usable instead of just aspirational. Right now it reads like a design doc, not executable guidance.

Other Things Worth Fixing

1. Description needs trigger phrases (Medium severity) – You're missing specific action verbs. Instead of "Curated repository of experiment hypotheses," try: "Capture experiment hypotheses with structured metadata, track status from idea to archived, and retrieve past learnings. Use when 'log hypothesis', 'capture experiment idea', or 'search past experiments'." This helps users know when to invoke you.

2. Framework lacks implementation steps (Medium severity) – "Metadata Schema – hypothesis ID, theme, persona..." reads like a spec. Make it actionable: "Assign unique ID (format: YYYY-MM-HYP-###), select theme from [list], identify target persona..." Users need to know how, not just what.

3. Gerund forms in triggers (Medium severity) – Your "When to Use" section says "Capturing new experiment ideas" and "Referencing past wins"—use imperative instead: "Capture new experiment ideas" and "Reference past wins." Small change, huge readability win.

4. No validation loop (Medium severity) – You mention "review cadence" but never define how to validate hypothesis quality. Add a checklist: measurable metric? supporting data? owner assigned? success/failure criteria?

Quick Wins

• Add description trigger phrases (+2 points)
• Create the three template reference files (+7 points)
• Switch to imperative voice in "When to Use" (+2 points)
• Add validation checklist to Framework (+2 points)

That's roughly 13 points of improvement with focused changes. Focus on the templates first—that's your biggest bang for buck.

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.