sql-translation

I took a look at your sql-translation skill and wanted to share some thoughts.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 81/100, solid B territory. This is based on Anthropic's best practices for skill design. Your strongest area is Spec Compliance (12/15)—the frontmatter is clean and the naming is right. Where you're losing points: Progressive Disclosure Architecture (21/30) and Utility (15/20). You've got good fundamentals here, but there's meaningful room to push this into A territory.

What's Working Well

• Solid trigger terms: dbplyr, SQL translation, R-to-SQL—these are discoverable and specific enough that the right developers will find this skill
• Clear workflow structure: Seven numbered steps with a validation checklist at the end gives developers a concrete path forward
• Comprehensive translation patterns: You cover scalar, aggregate, and window function translations, which hits the main SQL translation use cases people actually need

The Big One: Missing Reference Files

Here's what's holding you back most—everything is crammed into a single 189-line SKILL.md file. For a skill this detailed, that's a lot of cognitive load in one place.

Why it matters: The grading rubric values Progressive Disclosure Architecture (30 points), and bundling everything together costs you 9 points here. When developers load your skill, they should get the essentials first, then progressively unfold complexity.

The fix: Create a references/ folder with two files:

• references/translation-patterns.md — Move your code examples and helper functions here
• references/testing-guide.md — Extract the testing section with full before/after examples

Keep the main SKILL.md lean (50-75 lines) with just workflow steps and entry points. This isn't busy work—it actually makes your skill easier to use because developers land on the core workflow first, then drill into patterns only when they need them.

Impact: +5 points (gets you to 86).

Other Things Worth Fixing

1. No concrete input/output pairs — You have code templates, but no "R: log10(x) → SQL: log(x)" examples showing actual translation results. Add 4-5 of these to the testing section. (+2 points)

2. External skill dependency without bundling — Step 1 references the sql-research skill, but doesn't bundle it. Either inline the critical research steps or create a references/research-checklist.md so developers aren't left hanging. (+3 points)

3. Checklist voice — Your end-of-skill checklist uses implied "you" ("Researched SQL syntax"). Switch to infinitive form: "Research SQL syntax", "Create research file". Minor but matters for consistency. (+1 point)

Quick Wins

If you implement these three things in order, you'd land around 91/100 (A territory):

1. Create references folder, split content, lean out SKILL.md
2. Add 4-5 concrete R→SQL translation examples with expected output
3. Inline or bundle the sql-research dependency, fix checklist voice

You've got the domain knowledge and the structure is sound. These are just architectural moves to help the skill 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.