Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tool for robust segmentation of 104+ anatomical structures in CT images. Uses nnUNet-based models for whole-body, organ, and bone segmentation. One-line CLI for comprehensive body-part segmentation.
.claude/skills/mkurman-total-segmentator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -61% | 0% |
TotalSegmentator segments 104+ anatomical structures in CT images using nnUNet-based models. Run full-body, organ, or bone segmentation with a single CLI command.
bashuv pip install TotalSegmentator
bash# Full body segmentation (all 104 structures) TotalSegmentator -i input_ct.nii.gz -o output_seg.nii.gz # Organ-only segmentation (liver, kidneys, spleen, etc.) TotalSegmentator -i input_ct.nii.gz -o organ_seg.nii.gz -ta organ # Appendicular bones TotalSegmentator -i input_ct.nii.gz -o bone_seg.nii.gz -ta appendicular_bones
pythonfrom totalsegmentator.python_api import totalsegmentator segmentation = totalsegmentator("input_ct.nii.gz", "output_seg.nii.gz")
TotalSegmentator -i input.nii.gz -o output.nii.gztotal (104 structures), organ, vertebra, ribs, appendicular_bones| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | 9,538 | 2,021 | -79% | 1 | 1 | 0% | 1,726 | 641 | -63% | 0 | 0 | — |
case-01 | fail→pass | 8,968 | 3,136 | -65% | 1 | 1 | 0% | 1,381 | 726 | -47% | 0 | 0 | — |
case-03 | fail→fail | 8,888 | 4,045 | -54% | 1 | 1 | 0% | 1,524 | 989 | -35% | 0 | 0 | — |
case-04 | pass→pass | 4,558 | 3,201 | -30% | 1 | 1 | 0% | 777 | 819 | +5% | 0 | 0 | — |
case-05 | fail→pass | 8,850 | 4,002 | -55% | 1 | 1 | 0% | 1,438 | 969 | -33% | 0 | 0 | — |
case-06 | fail→pass | 4,302 | 1,939 | -55% | 1 | 1 | 0% | 758 | 644 | -15% | 0 | 0 | — |
case-07 | pass→pass | 3,821 | 1,441 | -62% | 1 | 1 | 0% | 565 | 508 | -10% | 0 | 0 | — |
case-08 | pass→pass | 8,405 | 4,644 | -45% | 1 | 1 | 0% | 1,475 | 1,164 | -21% | 0 | 0 | — |
case-09 | pass→pass | 11,117 | 6,747 | -39% | 1 | 1 | 0% | 1,897 | 1,413 | -26% | 0 | 0 | — |
case-10 | pass→pass | 13,016 | 10,748 | -17% | 1 | 1 | 0% | 2,245 | 2,351 | +5% | 0 | 0 | — |
case-11 | fail→pass | 9,338 | 2,013 | -78% | 1 | 1 | 0% | 1,598 | 620 | -61% | 0 | 0 | — |
case-12 | fail→fail | 15,885 | 12,219 | -23% | 1 | 1 | 0% | 2,825 | 2,430 | -14% | 0 | 0 | — |
case-13 | fail→fail | 10,223 | 5,780 | -43% | 1 | 1 | 0% | 1,808 | 1,304 | -28% | 0 | 0 | — |
case-14 | fail→pass | 16,773 | 33,347 | +99% | 1 | 1 | 0% | 2,789 | 2,323 | -17% | 0 | 0 | — |
case-15 | fail→pass | 12,872 | 2,713 | -79% | 1 | 1 | 0% | 2,197 | 782 | -64% | 0 | 0 | — |
case-16 | pass→pass | 7,577 | 2,301 | -70% | 1 | 1 | 0% | 1,366 | 712 | -48% | 0 | 0 | — |
case-17 | pass→pass | 5,273 | 2,293 | -57% | 1 | 1 | 0% | 889 | 611 | -31% | 0 | 0 | — |
case-18 | pass→pass | 3,304 | 2,354 | -29% | 1 | 1 | 0% | 537 | 605 | +13% | 0 | 0 | — |
case-19 | pass→pass | 2,199 | 2,366 | +8% | 1 | 1 | 0% | 358 | 665 | +86% | 0 | 0 | — |
case-20 | pass→pass | 5,782 | 1,994 | -66% | 1 | 1 | 0% | 903 | 519 | -43% | 0 | 0 | — |
case-21 | pass→pass | 5,258 | 2,727 | -48% | 1 | 1 | 0% | 822 | 697 | -15% | 0 | 0 | — |
case-22 | fail→pass | 7,218 | 2,106 | -71% | 1 | 1 | 0% | 1,231 | 683 | -45% | 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 +36 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.