Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Take the single most severe domain/item judgment as the overall verdict, for RoB2 (3-value), ROBINS-I (5-value), or AMSTAR-2 (pre-filtered by critical-domain status before worst-case). Use this after domain-level-judgment (for RoB2/ROBINS-I) or quality-appraisal-checklist (for AMSTAR-2) has produced per-domain/item judgments — this SOP has two structurally distinct upstream callers and must identify which value domain it received before applying the matching lookup rule. QUADAS-2 never reaches t
.claude/skills/yogsoth-ai-worst-case-lookup/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-23 | ✗→✓ | ▲ Improved | -73% | 0% |
Overall verdict = most severe domain/item value, on the caller's own scale. Merges what were originally 3 separate SOPs (RoB2-aggregate, ROBINS-I-aggregate, AMSTAR-2-aggregate) per coverage-audit M10's finding that they share one algorithm (worst-case-taking) differing only in value domain and, for AMSTAR-2, an extra pre-filter step — the same parameterization principle already used for unit-classification's label_set parameter, applied consistently here.
Subagent — spawned via spawn-agent skill.
Unlike most SOPs in this package, this one is called from two different places in the graph with two different input shapes (domain_judgments vs checklist_result). The prompt's Step 1 instruction to identify which was received before proceeding is load-bearing — applying RoB2's worst-case rule to AMSTAR-2's input (or vice versa) silently produces a wrong answer, not an error.
<!-- 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-01 | fail→fail | 13,587 | 31,317 | +130% | 1 | 1 | 0% | 1,667 | 6,244 | +275% | 0 | 0 | — |
case-02 | fail→fail | 9,638 | 9,316 | -3% | 1 | 1 | 0% | 854 | 1,081 | +27% | 0 | 0 | — |
case-03 | fail→fail | 9,924 | 9,324 | -6% | 1 | 1 | 0% | 896 | 1,074 | +20% | 0 | 0 | — |
case-04 | fail→pass | 10,355 | 11,862 | +15% | 1 | 1 | 0% | 1,002 | 1,591 | +59% | 0 | 0 | — |
case-05 | pass→pass | 9,672 | 18,573 | +92% | 1 | 1 | 0% | 833 | 3,217 | +286% | 0 | 0 | — |
case-06 | pass→pass | 10,313 | 12,722 | +23% | 1 | 1 | 0% | 1,052 | 1,808 | +72% | 0 | 0 | — |
case-07 | pass→pass | 9,511 | 15,435 | +62% | 1 | 1 | 0% | 855 | 2,391 | +180% | 0 | 0 | — |
case-08 | pass→pass | 9,831 | 9,468 | -4% | 1 | 1 | 0% | 810 | 974 | +20% | 0 | 0 | — |
case-09 | pass→pass | 17,624 | 23,396 | +33% | 1 | 1 | 0% | 2,744 | 4,457 | +62% | 0 | 0 | — |
case-10 | pass→pass | 19,717 | 20,011 | +1% | 1 | 1 | 0% | 2,810 | 3,263 | +16% | 0 | 0 | — |
case-11 | pass→pass | 15,631 | 18,895 | +21% | 1 | 1 | 0% | 1,783 | 3,020 | +69% | 0 | 0 | — |
case-12 | pass→pass | 19,519 | 20,695 | +6% | 1 | 1 | 0% | 2,822 | 3,764 | +33% | 0 | 0 | — |
case-13 | fail→fail | 37,377 | 19,959 | -47% | 1 | 1 | 0% | 2,287 | 1,425 | -38% | 0 | 0 | — |
case-14 | pass→pass | 10,051 | 12,754 | +27% | 1 | 1 | 0% | 852 | 1,687 | +98% | 0 | 0 | — |
case-15 | pass→pass | 10,553 | 11,154 | +6% | 1 | 1 | 0% | 1,064 | 1,528 | +44% | 0 | 0 | — |
case-16 | pass→pass | 8,746 | 10,153 | +16% | 1 | 1 | 0% | 678 | 1,267 | +87% | 0 | 0 | — |
case-17 | pass→pass | 11,064 | 9,148 | -17% | 1 | 1 | 0% | 911 | 958 | +5% | 0 | 0 | — |
case-18 | pass→pass | 11,436 | 19,792 | +73% | 1 | 1 | 0% | 1,246 | 3,167 | +154% | 0 | 0 | — |
case-19 | fail→pass | 18,905 | 11,133 | -41% | 1 | 1 | 0% | 2,128 | 1,362 | -36% | 0 | 0 | — |
case-20 | fail→pass | 18,228 | 18,772 | +3% | 1 | 1 | 0% | 2,267 | 2,673 | +18% | 0 | 0 | — |
case-21 | fail→pass | 19,822 | 10,115 | -49% | 1 | 1 | 0% | 2,568 | 1,191 | -54% | 0 | 0 | — |
case-22 | pass→pass | 10,405 | 11,029 | +6% | 1 | 1 | 0% | 1,006 | 1,619 | +61% | 0 | 0 | — |
case-23 | fail→pass | 26,630 | 8,822 | -67% | 1 | 1 | 0% | 3,585 | 962 | -73% | 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. The headline lift of +22 percentage points is the difference between those two pass rates over the 23 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.