Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build input spaces for two source concepts
.claude/skills/yogsoth-ai-input-space-construction/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✓→✗ | ▼ Worse | -58% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 81% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 83% | 0% |
| case-01 | ✗→✗ | ＝ Same ✗ | 38% | 0% |
| case-02 | ✗→✗ | ＝ Same ✗ | 24% | 0% |
Build input spaces for two source concepts, elaborating each into its elements, relations, and attributes.
Subagent — spawned via subagent-spawning/spawn-agent skill.
Input space construction requires deep elaboration of each concept's internal structure, identifying non-obvious elements and relations. Benefits from dedicated focused attention on each concept independently.
<!-- 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 | 64,737 | 61,608 | -5% | 1 | 1 | 0% | 2,476 | 3,412 | +38% | 0 | 0 | — |
case-02 | fail→fail | 25,367 | 98,224 | +287% | 1 | 1 | 0% | 3,212 | 3,969 | +24% | 0 | 0 | — |
case-03 | pass→fail | 25,710 | 38,007 | +48% | 1 | 1 | 0% | 3,176 | 1,348 | -58% | 0 | 0 | — |
case-04 | pass→pass | 24,638 | 33,254 | +35% | 1 | 1 | 0% | 2,364 | 4,285 | +81% | 0 | 0 | — |
case-05 | pass→pass | 21,102 | 33,426 | +58% | 1 | 1 | 0% | 2,504 | 4,584 | +83% | 0 | 0 | — |
case-06 | fail→fail | 10,346 | 25,099 | +143% | 1 | 1 | 0% | 1,571 | 1,797 | +14% | 0 | 0 | — |
case-07 | fail→fail | 13,974 | 11,027 | -21% | 1 | 1 | 0% | 985 | 870 | -12% | 0 | 0 | — |
case-08 | fail→fail | 15,062 | 59,625 | +296% | 1 | 1 | 0% | 2,251 | 4,258 | +89% | 0 | 0 | — |
case-09 | fail→fail | 10,690 | 8,412 | -21% | 1 | 1 | 0% | 766 | 1,281 | +67% | 0 | 0 | — |
case-10 | fail→fail | 37,594 | 19,190 | -49% | 1 | 1 | 0% | 4,646 | 3,748 | -19% | 0 | 0 | — |
case-11 | fail→fail | 23,754 | 28,307 | +19% | 1 | 1 | 0% | 4,150 | 3,881 | -6% | 0 | 0 | — |
case-12 | fail→fail | 26,504 | 27,154 | +2% | 1 | 1 | 0% | 3,902 | 5,471 | +40% | 0 | 0 | — |
case-13 | fail→fail | 13,939 | 43,069 | +209% | 1 | 1 | 0% | 1,721 | 910 | -47% | 0 | 0 | — |
case-14 | fail→fail | 23,251 | 18,819 | -19% | 1 | 1 | 0% | 3,767 | 2,432 | -35% | 0 | 0 | — |
case-15 | fail→fail | 18,481 | 23,191 | +25% | 1 | 1 | 0% | 3,642 | 3,276 | -10% | 0 | 0 | — |
case-16 | fail→fail | 16,373 | 12,719 | -22% | 1 | 1 | 0% | 1,945 | 1,895 | -3% | 0 | 0 | — |
case-17 | fail→fail | 28,885 | 36,355 | +26% | 1 | 1 | 0% | 5,429 | 5,778 | +6% | 0 | 0 | — |
case-18 | fail→fail | 7,986 | 32,947 | +313% | 1 | 1 | 0% | 1,316 | 4,216 | +220% | 0 | 0 | — |
case-19 | fail→fail | 22,376 | 45,379 | +103% | 1 | 1 | 0% | 3,680 | 7,324 | +99% | 0 | 0 | — |
case-20 | fail→fail | 11,208 | 26,271 | +134% | 1 | 1 | 0% | 2,203 | 3,013 | +37% | 0 | 0 | — |
case-21 | fail→fail | 9,932 | 13,883 | +40% | 1 | 1 | 0% | 1,654 | 2,372 | +43% | 0 | 0 | — |
case-22 | fail→fail | 17,683 | 26,336 | +49% | 1 | 1 | 0% | 2,965 | 3,875 | +31% | 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 20 counted toward the lift figure. The other 2 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 -5 percentage points is the difference between those two pass rates over the 20 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Other measured skills in the registry, with their headline benchmark lift.