Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Split a paper's text into sentence- or clause-level units (with character offsets) for downstream classification, at a caller-specified granularity and scope (full text, abstract-only, or intro-only). Use this as the mandatory first step whenever any sentence/clause-level classification method (Argumentative Zoning, CoreSC, PubMed-RCT, CSAbstruct, Swales move analysis, CODA-19) needs its input pre-segmented — always precedes unit-classification.
.claude/skills/yogsoth-ai-unit-segmentation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-20 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -75% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -32% | 0% |
Splits text into labeling units (sentence or clause granularity, scoped to full text/abstract/intro) — pure segmentation, no labeling.
Subagent — spawned via spawn-agent skill.
7 different classification methods (AZ, CoreSC, PubMed-RCT, NICTA-PIBOSO, CSAbstruct, CODA-19, Swales) all need pre-segmented units but disagree on granularity and scope — factoring segmentation out once, parameterized, avoids duplicating this logic inside unit-classification seven times over (graph correction L17/L18: the original graph was missing this step entirely, silently assuming pre-segmented input existed).
<!-- BEGIN available-tables (generated) -->
| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→pass | 20,647 | 11,608 | -44% | 1 | 1 | 0% | 2,368 | 1,261 | -47% | 0 | 0 | — |
case-01 | fail→pass | 23,325 | 16,536 | -29% | 1 | 1 | 0% | 2,841 | 2,006 | -29% | 0 | 0 | — |
case-02 | fail→pass | 17,628 | 6,755 | -62% | 1 | 1 | 0% | 1,982 | 493 | -75% | 0 | 0 | — |
case-03 | fail→pass | 20,960 | 14,174 | -32% | 1 | 1 | 0% | 2,360 | 1,678 | -29% | 0 | 0 | — |
case-04 | pass→pass | 16,663 | 9,828 | -41% | 1 | 1 | 0% | 1,839 | 1,052 | -43% | 0 | 0 | — |
case-05 | fail→pass | 19,657 | 12,756 | -35% | 1 | 1 | 0% | 2,169 | 1,476 | -32% | 0 | 0 | — |
case-06 | fail→pass | 16,178 | 7,615 | -53% | 1 | 1 | 0% | 1,670 | 674 | -60% | 0 | 0 | — |
case-07 | fail→pass | 18,277 | 7,347 | -60% | 1 | 1 | 0% | 1,978 | 529 | -73% | 0 | 0 | — |
case-08 | pass→fail | 13,279 | 8,651 | -35% | 1 | 1 | 0% | 1,281 | 688 | -46% | 0 | 0 | — |
case-09 | pass→pass | 19,985 | 11,653 | -42% | 1 | 1 | 0% | 2,189 | 1,239 | -43% | 0 | 0 | — |
case-10 | pass→pass | 15,706 | 8,225 | -48% | 1 | 1 | 0% | 1,604 | 735 | -54% | 0 | 0 | — |
case-11 | pass→pass | 24,341 | 11,788 | -52% | 1 | 1 | 0% | 2,929 | 1,389 | -53% | 0 | 0 | — |
case-12 | pass→pass | 23,982 | 8,879 | -63% | 1 | 1 | 0% | 2,873 | 840 | -71% | 0 | 0 | — |
case-13 | fail→fail | 18,442 | 8,220 | -55% | 1 | 1 | 0% | 2,113 | 738 | -65% | 0 | 0 | — |
case-14 | pass→pass | 21,087 | 10,086 | -52% | 1 | 1 | 0% | 2,450 | 1,004 | -59% | 0 | 0 | — |
case-15 | fail→pass | 20,140 | 8,563 | -57% | 1 | 1 | 0% | 2,283 | 795 | -65% | 0 | 0 | — |
case-16 | pass→pass | 12,320 | 8,326 | -32% | 1 | 1 | 0% | 1,211 | 735 | -39% | 0 | 0 | — |
case-17 | fail→pass | 19,766 | 13,173 | -33% | 1 | 1 | 0% | 2,578 | 1,695 | -34% | 0 | 0 | — |
case-18 | pass→pass | 18,970 | 15,646 | -18% | 1 | 1 | 0% | 2,221 | 1,783 | -20% | 0 | 0 | — |
case-19 | fail→pass | 17,525 | 12,096 | -31% | 1 | 1 | 0% | 2,024 | 1,204 | -41% | 0 | 0 | — |
case-21 | pass→pass | 17,529 | 14,183 | -19% | 1 | 1 | 0% | 1,793 | 1,580 | -12% | 0 | 0 | — |
case-22 | pass→pass | 10,964 | 7,996 | -27% | 1 | 1 | 0% | 902 | 665 | -26% | 0 | 0 | — |
case-23 | pass→pass | 9,320 | 11,593 | +24% | 1 | 1 | 0% | 827 | 1,306 | +58% | 0 | 0 | — |
case-24 | pass→pass | 9,138 | 10,019 | +10% | 1 | 1 | 0% | 736 | 1,117 | +52% | 0 | 0 | — |
case-25 | pass→pass | 13,397 | 11,432 | -15% | 1 | 1 | 0% | 1,507 | 1,365 | -9% | 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. 25 cases were attempted. The headline lift of +36 percentage points is the difference between those two pass rates over the 25 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.