tech-stack-evaluator

I just looked at how you're handling tech stack evaluation—there's a solid foundation here, but I'm noticing some architectural decisions are holding back what could be a really useful skill. What made you choose to pack everything into a single 430-line file?

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 77/100, which is solid C-grade territory. This is graded against Anthropic's skill architecture best practices—specifically around token efficiency, layered structure, and spec compliance. Your strongest area is Utility (16/20)—the framework itself is genuinely comprehensive with TCO analysis and security assessment built in. The weakest is Spec Compliance (11/15), mainly because the description lacks trigger phrases that help users discover and activate your skill.

What's Working Well

• Comprehensive evaluation framework: Your skill actually addresses a real gap. The multi-factor analysis (TCO, security, migration paths, team skill matching) is exactly what teams need when making stack decisions.
• Clear output structure: You've thoughtfully designed five different report formats (Executive Summary, Detailed Analysis, Comparison Matrix, etc.), which shows you understand different user needs.
• Good practical constraints: The skill boundaries are explicit—you handle "current popular stacks" and acknowledge limitations around bleeding-edge tech, which is honest and helpful.
• Solid examples: Your input examples show real scenarios (comparing React vs Vue, migrations from Angular), making it discoverable and immediately useful.

The Big One: Architecture is Fighting You

Your single-file design is the core issue. Everything—metrics calculations, best practices, limitations, report formats—is crammed into SKILL.md, which kills your token efficiency and violates progressive disclosure architecture.

Why it matters: Skills are supposed to be lazy-loaded. You want users to quickly understand what your skill does, then gradually discover how it works. Right now, they're getting hit with 430 lines of everything at once.

The fix: Create a references/ directory:

• references/metrics_calculations.md - Move lines 66-141 (capability details and calculations)
• references/report_formats.md - Move output format specs (lines 73-145)
• references/best_practices.md - Move the Best Practices section (lines 353-384)

Keep SKILL.md to ~150-200 lines as a clean entry point. This alone gains you +7 points (goes from 18→25 on PDA).

Other Things Worth Fixing

1. Add trigger phrases to the description (11→13 on Spec Compliance). Currently: "Comprehensive technology stack evaluation..." Instead: "Use when asked to 'compare tech stacks', 'evaluate react vs vue', 'tco analysis', 'migration from angular', 'cloud provider comparison'." This is a quick +2 point win.

2. Cut the repetition (lines 10-22 and 66-141 say basically the same thing). You're burning ~80 tokens describing capabilities twice. Move one to references, reference it in SKILL.md. Gains +3 points on token economy.

3. Watch the voice consistency (Best Practices section uses "you/your"—should be imperative: "Define clear use cases" instead of "Define your clear use cases"). Small fix, +1 point.

4. Add actual output examples—The report formats are well-structured, but showing an actual sample Executive Summary or Comparison Matrix would make Utility jump from 16→18.

Quick Wins (Most Impact First)

• ✅ Create references/ subdirectory structure (+7 points)
• ✅ Add trigger phrases to frontmatter description (+2 points)
• ✅ Remove Capabilities repetition, keep one version (+3 points)
• ✅ Add 1-2 example output reports (+2 points)

You're looking at +14 points (77→91) with these changes. The skill is genuinely useful—it just needs to breathe a little.

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.