Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Add and manage evaluation results in Hugging Face model cards. Supports extracting eval tables from README content, importing scores from Artificial Analysis API, and running custom model evaluations with vLLM/lighteval. Works with the model-index metadata format.
.claude/skills/sickn33-hugging-face-evaluation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 162% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 226% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 226% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 207% | 0% |
This skill provides tools to add structured evaluation results to Hugging Face model cards. It supports multiple methods for adding evaluation data:
Read the detailed guide before executing this skill. It retains the complete procedure and reference material. Treat its safety, prerequisites, and validation requirements as mandatory. For focused work, load the relevant sections; for end-to-end work, read the guide completely.
model-index metadata for a model release.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 16,194 | 66,286 | +309% | 1 | 1 | 0% | 3,293 | 6,872 | +109% | 0 | 0 | — |
case-02 | fail→fail | 43,034 | 7,777 | -82% | 1 | 1 | 0% | 2,430 | 6,912 | +184% | 0 | 0 | — |
case-03 | fail→fail | 34,784 | 41,856 | +20% | 1 | 1 | 0% | 714 | 7,723 | +982% | 0 | 0 | — |
case-04 | pass→pass | 18,666 | 17,044 | -9% | 1 | 1 | 0% | 3,452 | 9,861 | +186% | 0 | 0 | — |
case-05 | pass→pass | 23,107 | 16,824 | -27% | 1 | 1 | 0% | 3,588 | 9,909 | +176% | 0 | 0 | — |
case-06 | pass→pass | 10,299 | 7,917 | -23% | 1 | 1 | 0% | 1,799 | 7,895 | +339% | 0 | 0 | — |
case-07 | fail→pass | 16,818 | 5,337 | -68% | 1 | 1 | 0% | 2,856 | 7,472 | +162% | 0 | 0 | — |
case-08 | fail→pass | 13,400 | 5,413 | -60% | 1 | 1 | 0% | 2,314 | 7,534 | +226% | 0 | 0 | — |
case-09 | pass→pass | 14,891 | 5,801 | -61% | 1 | 1 | 0% | 2,458 | 7,554 | +207% | 0 | 0 | — |
case-10 | fail→pass | 22,046 | 3,819 | -83% | 1 | 1 | 0% | 4,116 | 7,153 | +74% | 0 | 0 | — |
case-11 | fail→pass | 13,082 | 5,317 | -59% | 1 | 1 | 0% | 2,307 | 7,530 | +226% | 0 | 0 | — |
case-12 | fail→pass | 13,622 | 4,657 | -66% | 1 | 1 | 0% | 2,414 | 7,412 | +207% | 0 | 0 | — |
case-13 | fail→pass | 15,247 | 7,791 | -49% | 1 | 1 | 0% | 2,765 | 8,133 | +194% | 0 | 0 | — |
case-14 | fail→pass | 11,398 | 2,907 | -74% | 1 | 1 | 0% | 1,991 | 6,987 | +251% | 0 | 0 | — |
case-15 | fail→pass | 10,725 | 3,613 | -66% | 1 | 1 | 0% | 1,912 | 7,168 | +275% | 0 | 0 | — |
case-16 | fail→pass | 9,411 | 4,743 | -50% | 1 | 1 | 0% | 1,565 | 7,411 | +374% | 0 | 0 | — |
case-17 | fail→fail | 14,566 | 4,256 | -71% | 1 | 1 | 0% | 2,768 | 7,149 | +158% | 0 | 0 | — |
case-18 | fail→pass | 9,331 | 2,512 | -73% | 1 | 1 | 0% | 1,499 | 7,001 | +367% | 0 | 0 | — |
case-19 | pass→pass | 8,744 | 2,760 | -68% | 1 | 1 | 0% | 1,455 | 6,961 | +378% | 0 | 0 | — |
case-20 | fail→pass | 14,290 | 3,333 | -77% | 1 | 1 | 0% | 2,179 | 7,093 | +226% | 0 | 0 | — |
case-21 | fail→pass | 11,935 | 3,978 | -67% | 1 | 1 | 0% | 1,914 | 7,287 | +281% | 0 | 0 | — |
case-22 | fail→pass | 15,112 | 5,575 | -63% | 1 | 1 | 0% | 2,571 | 7,620 | +196% | 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, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +59 percentage points is the difference between those two pass rates over the 19 comparable cases.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.