Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →SOP for creating a claim page in the vault — atomic proposition with type classification, source attribution, and initial confidence.
.claude/skills/yogsoth-ai-claim-page-creation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 106% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 167% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 21% | 0% |
Create a claim page in the vault representing a single atomic proposition.
CC file write + vault_add_edge
wiki/claims/<claim-slug>.md with frontmatter:type: claimclaim-type: one of empirical, definitional, causal, normative, existential]confidence: initial confidence (0.0-1.0) based on source qualitysource: attribution to the originating source pagederived_from (claim → source page)related_to edges[[dir/slug]] pointing to the target (dir/slug = target path minus .md). Place inline at semantically relevant location. Skip if already present.<HARD-GATE> Claim must be atomic — one proposition only. If it contains "and", "but", "because", or multiple independent assertions, it must be decomposed first. </HARD-GATE>
Returns: { path: string, claim_type: string, confidence: number }
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,510 | 13,128 | +138% | 1 | 1 | 0% | 1,096 | 1,772 | +62% | 0 | 0 | — |
case-02 | fail→fail | 7,633 | 11,136 | +46% | 1 | 1 | 0% | 1,611 | 945 | -41% | 0 | 0 | — |
case-07 | pass→fail | 15,487 | 8,649 | -44% | 1 | 1 | 0% | 3,177 | 1,277 | -60% | 0 | 0 | — |
case-03 | fail→pass | 8,908 | 9,226 | +4% | 1 | 1 | 0% | 1,699 | 2,058 | +21% | 0 | 0 | — |
case-04 | pass→pass | 5,396 | 3,024 | -44% | 1 | 1 | 0% | 933 | 836 | -10% | 0 | 0 | — |
case-05 | pass→pass | 4,898 | 5,386 | +10% | 1 | 1 | 0% | 817 | 1,143 | +40% | 0 | 0 | — |
case-06 | pass→pass | 5,632 | 3,525 | -37% | 1 | 1 | 0% | 1,051 | 886 | -16% | 0 | 0 | — |
case-08 | fail→fail | 10,954 | 12,061 | +10% | 1 | 1 | 0% | 2,042 | 2,812 | +38% | 0 | 0 | — |
case-09 | fail→fail | 4,980 | 7,733 | +55% | 1 | 1 | 0% | 861 | 814 | -5% | 0 | 0 | — |
case-10 | pass→fail | 8,936 | 9,037 | +1% | 1 | 1 | 0% | 1,560 | 1,051 | -33% | 0 | 0 | — |
case-11 | fail→pass | 8,319 | 10,405 | +25% | 1 | 1 | 0% | 1,504 | 2,055 | +37% | 0 | 0 | — |
case-12 | fail→pass | 5,476 | 10,258 | +87% | 1 | 1 | 0% | 1,057 | 2,173 | +106% | 0 | 0 | — |
case-13 | fail→fail | 8,308 | 10,711 | +29% | 1 | 1 | 0% | 1,644 | 2,339 | +42% | 0 | 0 | — |
case-14 | fail→fail | 9,489 | 12,989 | +37% | 1 | 1 | 0% | 1,776 | 1,110 | -38% | 0 | 0 | — |
case-15 | fail→pass | 3,608 | 14,513 | +302% | 1 | 1 | 0% | 736 | 1,966 | +167% | 0 | 0 | — |
case-16 | fail→fail | 4,608 | 7,987 | +73% | 1 | 1 | 0% | 830 | 855 | +3% | 0 | 0 | — |
case-17 | pass→pass | 5,220 | 12,980 | +149% | 1 | 1 | 0% | 984 | 2,981 | +203% | 0 | 0 | — |
case-18 | pass→fail | 8,691 | 8,226 | -5% | 1 | 1 | 0% | 1,630 | 706 | -57% | 0 | 0 | — |
case-19 | fail→fail | 5,374 | 11,706 | +118% | 1 | 1 | 0% | 259 | 1,252 | +383% | 0 | 0 | — |
case-20 | fail→fail | 7,735 | 7,725 | -0% | 1 | 1 | 0% | 1,512 | 739 | -51% | 0 | 0 | — |
case-21 | fail→fail | 3,799 | 12,978 | +242% | 1 | 1 | 0% | 713 | 1,285 | +80% | 0 | 0 | — |
case-22 | fail→pass | 7,989 | 6,514 | -18% | 1 | 1 | 0% | 1,329 | 1,603 | +21% | 0 | 0 | — |
case-23 | fail→fail | 3,144 | 6,350 | +102% | 1 | 1 | 0% | 462 | 676 | +46% | 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. 23 cases were attempted, and 11 counted toward the lift figure. The other 12 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 +9 percentage points is the difference between those two pass rates over the 11 comparable cases. 5 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.