Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Strategy for extracting claims from source material — identify propositions, decompose compound claims, classify claim types, create claim pages in the vault.
.claude/skills/yogsoth-ai-claim-extraction/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 528% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 689% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 329% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -49% | 0% |
Identify and extract claims from source material. Decomposes compound statements into atomic propositions, classifies claim types, and creates structured claim pages in the vault.
| Metric | Small | Medium | Large | |--------|-------|--------|-------| | Sources processed | 5 | 12 | 25 | | Claims extracted | 10 | 25 | 50 | | Compound claims decomposed | 3 | 8 | 15 |
<HARD-GATE> Print before every iteration:
| Metric | Target | Current | % | |--------|--------|---------|---| | Sources processed | — | — | — | | Claims extracted | — | — | — | | Compound claims decomposed | — | — | — | </HARD-GATE>
Cannot exit until 80% of budget targets met.
After budget gate passes, self-check:
Max 2 extra iterations if gaps found.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | knowledge-structuring-claim-decomposition | Tactic for decomposing compound claims into atomic propositions — identify logical structure, separate conjunctions, extract implicit assumptions. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | claim-page-creation | SOP for creating a claim page in the vault — atomic proposition with type classification, source attribution, and initial confidence. | | rebuttal-documentation | SOP for documenting rebuttals and counter-claims — create rebuttal pages with typed contradiction edges and source attribution. |
<!-- 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-01 | fail→pass | 5,322 | 23,401 | +340% | 1 | 1 | 0% | 795 | 4,995 | +528% | 0 | 0 | — |
case-02 | fail→pass | 4,843 | 31,538 | +551% | 1 | 1 | 0% | 833 | 6,570 | +689% | 0 | 0 | — |
case-03 | fail→pass | 7,054 | 24,093 | +242% | 1 | 1 | 0% | 1,209 | 5,186 | +329% | 0 | 0 | — |
case-04 | fail→pass | 10,651 | 4,339 | -59% | 1 | 1 | 0% | 1,873 | 1,250 | -33% | 0 | 0 | — |
case-05 | fail→pass | 13,317 | 3,226 | -76% | 1 | 1 | 0% | 2,177 | 1,119 | -49% | 0 | 0 | — |
case-06 | fail→pass | 17,531 | 4,503 | -74% | 1 | 1 | 0% | 3,131 | 1,302 | -58% | 0 | 0 | — |
case-07 | fail→pass | 5,715 | 9,943 | +74% | 1 | 1 | 0% | 1,065 | 2,289 | +115% | 0 | 0 | — |
case-08 | fail→pass | 9,387 | 3,577 | -62% | 1 | 1 | 0% | 1,556 | 1,122 | -28% | 0 | 0 | — |
case-09 | fail→pass | 11,503 | 9,126 | -21% | 1 | 1 | 0% | 1,724 | 2,028 | +18% | 0 | 0 | — |
case-14 | pass→pass | 9,655 | 11,440 | +18% | 1 | 1 | 0% | 1,525 | 2,421 | +59% | 0 | 0 | — |
case-10 | fail→pass | 6,507 | 1,164 | -82% | 1 | 1 | 0% | 903 | 688 | -24% | 0 | 0 | — |
case-11 | pass→pass | 15,979 | 10,782 | -33% | 1 | 1 | 0% | 2,649 | 2,528 | -5% | 0 | 0 | — |
case-12 | fail→pass | 16,401 | 12,772 | -22% | 1 | 1 | 0% | 2,750 | 2,812 | +2% | 0 | 0 | — |
case-13 | pass→pass | 7,125 | 5,123 | -28% | 1 | 1 | 0% | 1,395 | 1,506 | +8% | 0 | 0 | — |
case-15 | pass→pass | 6,135 | 2,668 | -57% | 1 | 1 | 0% | 1,098 | 987 | -10% | 0 | 0 | — |
case-16 | pass→pass | 12,373 | 4,712 | -62% | 1 | 1 | 0% | 2,181 | 1,329 | -39% | 0 | 0 | — |
case-17 | pass→pass | 3,578 | 10,216 | +186% | 1 | 1 | 0% | 679 | 2,317 | +241% | 0 | 0 | — |
case-18 | fail→pass | 12,721 | 8,977 | -29% | 1 | 1 | 0% | 2,220 | 1,927 | -13% | 0 | 0 | — |
case-19 | fail→pass | 21,131 | 1,872 | -91% | 1 | 1 | 0% | 1,331 | 825 | -38% | 0 | 0 | — |
case-20 | fail→fail | 5,534 | 6,876 | +24% | 1 | 1 | 0% | 904 | 1,699 | +88% | 0 | 0 | — |
case-21 | fail→fail | 2,781 | 6,766 | +143% | 1 | 1 | 0% | 409 | 1,753 | +329% | 0 | 0 | — |
case-22 | fail→fail | 5,866 | 8,402 | +43% | 1 | 1 | 0% | 1,084 | 1,965 | +81% | 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 21 counted toward the lift figure. The other 1 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 21 comparable cases. 2 cases got worse with the skill loaded, and they are 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.