Free Tool · No Signup Required
Hermes Skill Quality Checker
Paste a Hermes skill file. Get an instant quality score across five dimensions. 100% browser-side — nothing leaves your machine.
0 / 8,000 chars · ~0 tokens
Paste a skill file and click Check Skill
Scored across 5 dimensions: readability, specificity, structure, error handling, and token efficiency.
How the score is calculated
Each dimension scores 0–20. Total is out of 100.
Readability (0–20)
Measures average sentence length and vague qualifier density. Short, direct imperatives score highest. Hedge words like "might", "generally", "could" reduce the score — they create ambiguity for the model.
Specificity (0–20)
Counts code blocks, inline `commands`, file paths, and API references. Abstract prose with no concrete anchors scores low. Every exact parameter, field name, or command helps the model execute reliably.
Structure (0–20)
Numbered steps score highest — Hermes follows them sequentially. Bullet points are good. Section headings add bonus points. Pure prose with no lists scores near zero.
Error Handling (0–20)
Counts explicit failure instructions: retry logic, empty-result handling, invalid input checks. A dedicated error-handling section earns a bonus. Without it, Hermes may loop or return wrong data silently.
Token Efficiency (0–20)
Estimates tokens (chars ÷ 4). Skills under 600 tokens score 15–20. Skills over 1,800 tokens score near zero — they consume context window that task data needs. Split long skills into focused sub-skills.
Shareable results
After scoring, click Share to copy a URL containing your full skill file encoded in the URL hash. Anyone with the link sees the same score. No account, no server — everything runs in the browser.
Run this check on all your Hermes skills — automatically
BenchGen scores your entire skill library, tracks quality over time, and alerts you when a model switch causes regressions. Free tier available.
Learn about BenchGen for Hermes →