tech-stack-evaluator

I've been diving into how developers choose their tech stacks, and your skill's approach to systematizing that decision-making really caught my eye—though the grading suggests there's some untapped potential in how you're structuring the guidance.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 76/100, solidly C territory. This evaluation is based on Anthropic's Progressive Disclosure Architecture standards—the five pillars that make skills actually useful to developers. Your strongest area is Utility (16/20)—the core capability is genuinely solid. But PDA is your weak point (18/30)—you're cramming 430 lines into one file when you should be layering the complexity.

What's Working Well

• Comprehensive coverage: You're addressing a real gap. The TCO analysis, security assessment, and migration path evaluation? That's exactly what teams need when evaluating stacks.
• Clear metadata: Your triggers are specific ("tech stack evaluator", "run tech stack evaluator") and your description paints a picture of what this does. That's better than most.
• Solid utility design: You've built in weighted criteria and progressive disclosure concepts—users can select analysis sections rather than getting dumped on with everything. That's thoughtful.
• Structured sections: The progression from Use Cases through Output Formats feels logical. Navigation isn't broken, it's just... dense.

The Big One: Your Architecture is Upside Down

Here's the thing: all 430 lines live in one SKILL.md file. That's the killer for PDA. You should have:

• SKILL.md: Lightweight entry point (~80 lines max)
• references/metrics.md: All your metric calculations (currently ~180 lines of repetitive detail)
• references/examples.md: Concrete input/output pairs showing exactly what good results look like
• references/workflows.md: Step-by-step walkthroughs for common scenarios

This isn't just tidying up—it moves you from 18/30 to ~26/30 on PDA (+8 points easily). Developers skim the main file to decide if they want to use you, then dig into references if they need details. Right now they're wading through everything upfront.

Other Things Worth Fixing

1. Add a Table of Contents to SKILL.md right after the metadata. For a 430-line file, this is mandatory. Adds ~2 points (Navigation Signals).

2. Condense the Metrics section — lines 227-350 repeat the same concept definitions multiple times. Move detailed formulas to references/metrics.md and keep a summary table in the main file. This saves ~100 lines and improves token economy (+3 points).

3. Fix the voice — you're mixing descriptive language ("provides", "Users can") with imperative ("Define", "Provide"). Pick one and stick with it. Consistent infinitive/imperative throughout adds +2 points.

4. Show actual output examples — you describe the output format but don't show a concrete before/after. Add a real example in references/examples.md showing how a "quick comparison" output looks vs. a "detailed analysis". That's +2-3 points (Utility).

Quick Wins

• Highest ROI: Create the references/ directory structure. This alone gets you to ~84/100.
• Then: Add Table of Contents, condense Metrics section, move detailed content to references.
• Finally: Fix voice consistency and add concrete output examples.

You've got a genuinely useful skill here. The issue isn't capability—it's presentation architecture. Restructure for Progressive Disclosure and you're looking at 85+/100.

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.