hypothesis-generation

I took a look at your hypothesis-generation skill and wanted to share some thoughts.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 85/100, solid B territory. This is graded against Anthropic's skill best practices. Your strongest area is Progressive Disclosure Architecture (27/30) — the way you layered everything across SKILL.md plus three reference files is textbook stuff. The weakest spot is Spec Compliance (11/15), mainly because you're missing trigger phrases in the description.

What's Working Well

• Solid file architecture — Main SKILL.md at 156 lines, with ~1300 lines of detail tucked into references. That's the right balance. The 4-file structure keeps the main entry point clean without hiding depth.
• Comprehensive workflow — Your 8-step numbered process with sub-bullets and quality evaluation checklists is genuinely useful. Step 5 (quality assessment) does real validation work.
• Strong template game — That 303-line output template for hypothesis statements is practical and gives users a clear target. Reference files include worked examples for experimental designs.
• Consistent vocabulary — Terms like "testable", "falsifiable", and "mechanism" are used throughout, which makes this feel coherent.

The Big One: Add Trigger Phrases to Your Description

Right now your description is:

"Generate testable hypotheses. Formulate from observations, design experiments, explore competing exp..."

It reads like an abstract rather than actionable metadata. Trigger phrases are how Claude finds skills — they're the keywords someone searches for. You need explicit triggers like:

description: Performs hypothesis generation operations. Use when asked to "generate hypotheses", "design experiment", "test predictions", or "competing explanations". Works across biomedical, physics, and experimental research domains.

This fixes three things at once: tells Claude when to activate you, makes you discoverable, and gives context about your domain specificity. You'll pick up about +2 points here.

Other Things Worth Fixing

1. Add a table of contents to SKILL.md — At 156 lines, readers would benefit from a quick TOC linking to Workflow, Quality Standards, and Resources. Low effort, +1 point.

2. Trim the literature search file — Your references/literature_search_strategies.md is 506 lines and has some repetitive examples. Condense to ~250 by consolidating similar techniques and removing duplicate worked examples. Tightens PDA and gains +2.

3. Tighten the voice in "When to Use" — You've got "This skill should be used when..." which is passive. Switch to imperative: "Activate for:" or just "Use when:" and list triggers. Minor, but +1 toward writing style.

4. Make the description domain-specific — You mention "scientific inquiry across domains," which is true but vague. Try "for biomedical, experimental, and mechanistic research questions" instead. More specific, easier to discover.

Quick Wins

• Add trigger phrases to YAML frontmatter description (+2 points)
• Condense literature search reference by 250 lines (+2 points)
• Add TOC to SKILL.md (+1 point)
• Tighten voice in "When to Use" section (+1 point)

That's a potential 88-90/100 with focused edits. The structure is already there — this is polish work.

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.