Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when auditing URL structures for word separator format, generating slugs from titles, or planning a URL migration from underscores or spaces to hyphens with appropriate redirects.
.claude/skills/thedaviddias-hyphens/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | -14% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | 6% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 13% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -4% | 0% |
Google's search systems interpret hyphens in URLs as word boundaries, enabling it to match the URL against individual keyword queries. Underscores join words into a single token, meaning /seo_practices is seen as one word 'seopractices' rather than two separate words 'seo' and 'practices'.
-) to separate words in URLs — Google treats hyphens as word separators, underscores as part of the word/best-seo-practices is better than /best_seo_practices or /bestseopracticesInspect the URL path segment of each page. Flag any URL that contains: (1) Underscores (_) as word separators (e.g., /best_practices). (2) Spaces encoded as %20 or + in path segments. (3) Mixed separators in the same URL (e.g., /best-practices_guide). Report the count of non-hyphenated URLs by category.
_ with -, spaces/%20 with -.Google's John Mueller has confirmed that Google interprets hyphens as word separators in URLs, but treats underscores as word joiners. This means a URL with underscores matches fewer individual keyword queries. For example, /web_design would match 'web_design' as a single token, while /web-design matches both 'web design' and 'webdesign' queries.
Parse all URL paths used in the application's router or CMS slug fields. Flag any path segment containing _ characters between words or any %20 space encoding. Verify slug-generation utilities produce lowercase hyphenated output. Check redirect rules to confirm old underscore URLs are permanently (301) redirected to their hyphenated equivalents.
For full implementation details, code examples, and framework-specific guidance, see references/rule.md.
Rule page: https://frontendchecklist.io/en/rules/seo/hyphens
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 12,170 | 5,865 | -52% | 1 | 1 | 0% | 2,011 | 1,723 | -14% | 0 | 0 | — |
case-01 | pass→pass | 8,851 | 6,820 | -23% | 1 | 1 | 0% | 1,817 | 1,924 | +6% | 0 | 0 | — |
case-02 | pass→pass | 9,974 | 7,950 | -20% | 1 | 1 | 0% | 1,857 | 2,107 | +13% | 0 | 0 | — |
case-03 | pass→pass | 10,642 | 7,654 | -28% | 1 | 1 | 0% | 2,188 | 2,096 | -4% | 0 | 0 | — |
case-04 | pass→pass | 9,157 | 5,230 | -43% | 1 | 1 | 0% | 1,798 | 1,805 | +0% | 0 | 0 | — |
case-05 | pass→pass | 6,696 | 5,627 | -16% | 1 | 1 | 0% | 1,459 | 1,779 | +22% | 0 | 0 | — |
case-07 | pass→pass | 10,719 | 6,328 | -41% | 1 | 1 | 0% | 1,927 | 1,923 | -0% | 0 | 0 | — |
case-08 | pass→pass | 8,448 | 4,990 | -41% | 1 | 1 | 0% | 1,789 | 1,628 | -9% | 0 | 0 | — |
case-09 | pass→pass | 4,843 | 3,854 | -20% | 1 | 1 | 0% | 1,013 | 1,436 | +42% | 0 | 0 | — |
case-10 | pass→pass | 6,156 | 4,187 | -32% | 1 | 1 | 0% | 1,222 | 1,461 | +20% | 0 | 0 | — |
case-16 | pass→pass | 10,739 | 5,969 | -44% | 1 | 1 | 0% | 2,242 | 1,854 | -17% | 0 | 0 | — |
case-11 | pass→pass | 9,125 | 7,332 | -20% | 1 | 1 | 0% | 1,913 | 2,041 | +7% | 0 | 0 | — |
case-12 | fail→pass | 16,093 | 9,050 | -44% | 1 | 1 | 0% | 3,235 | 2,193 | -32% | 0 | 0 | — |
case-13 | pass→pass | 10,914 | 5,081 | -53% | 1 | 1 | 0% | 1,986 | 1,711 | -14% | 0 | 0 | — |
case-14 | pass→pass | 11,465 | 7,264 | -37% | 1 | 1 | 0% | 2,107 | 2,102 | -0% | 0 | 0 | — |
case-15 | pass→pass | 7,117 | 3,703 | -48% | 1 | 1 | 0% | 1,344 | 1,336 | -1% | 0 | 0 | — |
case-17 | pass→pass | 13,760 | 12,411 | -10% | 1 | 1 | 0% | 2,737 | 2,933 | +7% | 0 | 0 | — |
case-18 | pass→pass | 6,876 | 4,313 | -37% | 1 | 1 | 0% | 1,496 | 1,538 | +3% | 0 | 0 | — |
case-19 | pass→pass | 10,363 | 8,220 | -21% | 1 | 1 | 0% | 1,904 | 2,233 | +17% | 0 | 0 | — |
case-20 | pass→pass | 15,755 | 14,661 | -7% | 1 | 1 | 0% | 3,154 | 3,377 | +7% | 0 | 0 | — |
case-21 | pass→pass | 8,965 | 5,758 | -36% | 1 | 1 | 0% | 1,694 | 1,867 | +10% | 0 | 0 | — |
case-22 | pass→pass | 10,659 | 7,015 | -34% | 1 | 1 | 0% | 2,131 | 2,091 | -2% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.