Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Extract and steel-man the core arguments supporting a given opinion cluster.
.claude/skills/yogsoth-ai-argument-extraction/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✓→✓ | ＝ Same ✓ | -5% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -18% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 283% | 0% |
| case-01 | ✗→✗ | ＝ Same ✗ | -20% | 0% |
| case-02 | ✗→✗ | ＝ Same ✗ | 43% | 0% |
Extract the core arguments supporting a given opinion cluster and present them in their strongest possible form (steel-manned). Synthesizes reasoning from multiple perspectives within the cluster into coherent, well-structured arguments.
Spawn a subagent that takes a cluster characterization and the relevant judgments, then produces a set of steel-manned arguments representing the cluster's position.
Output MUST contain: at least 1 argument per cluster, each with claim, evidence, reasoning, and strength fields. Arguments must be steel-manned (strongest possible version).
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. Used by SOPs that declare execution: subagent. |
<!-- 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 | 15,978 | 10,994 | -31% | 1 | 1 | 0% | 2,498 | 1,993 | -20% | 0 | 0 | — |
case-02 | fail→fail | 17,342 | 23,908 | +38% | 1 | 1 | 0% | 2,828 | 4,057 | +43% | 0 | 0 | — |
case-03 | pass→pass | 11,977 | 10,615 | -11% | 1 | 1 | 0% | 2,025 | 1,932 | -5% | 0 | 0 | — |
case-04 | pass→pass | 18,145 | 13,639 | -25% | 1 | 1 | 0% | 3,244 | 2,644 | -18% | 0 | 0 | — |
case-05 | pass→pass | 4,836 | 16,524 | +242% | 1 | 1 | 0% | 822 | 3,146 | +283% | 0 | 0 | — |
case-06 | fail→fail | 4,656 | 5,738 | +23% | 1 | 1 | 0% | 770 | 1,065 | +38% | 0 | 0 | — |
case-07 | fail→fail | 7,062 | 14,924 | +111% | 1 | 1 | 0% | 1,173 | 2,801 | +139% | 0 | 0 | — |
case-08 | fail→fail | 7,185 | 11,867 | +65% | 1 | 1 | 0% | 1,072 | 2,247 | +110% | 0 | 0 | — |
case-09 | fail→fail | 7,770 | 11,873 | +53% | 1 | 1 | 0% | 1,237 | 2,055 | +66% | 0 | 0 | — |
case-10 | fail→fail | 5,741 | 8,695 | +51% | 1 | 1 | 0% | 972 | 1,690 | +74% | 0 | 0 | — |
case-11 | fail→fail | 10,715 | 18,610 | +74% | 1 | 1 | 0% | 1,575 | 3,186 | +102% | 0 | 0 | — |
case-12 | fail→fail | 7,555 | 11,544 | +53% | 1 | 1 | 0% | 1,164 | 2,019 | +73% | 0 | 0 | — |
case-13 | fail→fail | 10,765 | 13,746 | +28% | 1 | 1 | 0% | 1,671 | 2,340 | +40% | 0 | 0 | — |
case-14 | fail→fail | 9,045 | 12,771 | +41% | 1 | 1 | 0% | 1,386 | 2,176 | +57% | 0 | 0 | — |
case-15 | fail→fail | 9,689 | 13,682 | +41% | 1 | 1 | 0% | 1,478 | 2,327 | +57% | 0 | 0 | — |
case-16 | fail→fail | 11,119 | 14,102 | +27% | 1 | 1 | 0% | 1,810 | 2,320 | +28% | 0 | 0 | — |
case-17 | fail→fail | 9,966 | 15,279 | +53% | 1 | 1 | 0% | 1,552 | 2,633 | +70% | 0 | 0 | — |
case-18 | fail→fail | 6,474 | 13,798 | +113% | 1 | 1 | 0% | 992 | 2,242 | +126% | 0 | 0 | — |
case-19 | fail→fail | 8,996 | 7,642 | -15% | 1 | 1 | 0% | 1,441 | 1,459 | +1% | 0 | 0 | — |
case-20 | fail→fail | 5,075 | 12,181 | +140% | 1 | 1 | 0% | 847 | 2,147 | +153% | 0 | 0 | — |
case-21 | fail→fail | 9,427 | 13,258 | +41% | 1 | 1 | 0% | 1,439 | 2,411 | +68% | 0 | 0 | — |
case-22 | fail→fail | 6,766 | 12,989 | +92% | 1 | 1 | 0% | 1,066 | 2,161 | +103% | 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 0 percentage points is the difference between those two pass rates over the 22 comparable cases.
Other measured skills in the registry, with their headline benchmark lift.