Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Construct a hierarchical field map from paper collection — multi-level taxonomy with parent/child relationships, paper counts per node, and maturity indicators. Used by scoping-survey.
.claude/skills/yogsoth-ai-taxonomy-mapping/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | -80% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -53% | 0% |
| case-18 | ✓→✓ | ＝ Same ✓ | -64% | 0% |
| case-20 | ✓→✓ | ＝ Same ✓ | -10% | 0% |
| case-22 | ✓→✓ | ＝ Same ✓ | 43% | 0% |
Construct a hierarchical field map from a paper collection.
Subagent — spawned via subagent-spawning/spawn-agent skill.
Taxonomy construction requires holding the entire paper collection in context and iteratively refining hierarchical relationships. Dedicated context for this structural analysis.
<!-- 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-12 | fail→fail | 26,807 | 64,549 | +141% | 1 | 1 | 0% | 3,865 | 8,369 | +117% | 0 | 0 | — |
case-01 | fail→fail | 28,218 | 24,579 | -13% | 1 | 1 | 0% | 3,819 | 1,291 | -66% | 0 | 0 | — |
case-13 | fail→fail | 24,805 | 37,540 | +51% | 1 | 1 | 0% | 3,430 | 5,578 | +63% | 0 | 0 | — |
case-02 | fail→fail | 13,343 | 21,987 | +65% | 1 | 1 | 0% | 1,313 | 1,065 | -19% | 0 | 0 | — |
case-03 | fail→fail | 31,046 | 23,913 | -23% | 1 | 1 | 0% | 4,647 | 954 | -79% | 0 | 0 | — |
case-04 | fail→fail | 41,843 | 41,176 | -2% | 1 | 1 | 0% | 6,874 | 7,513 | +9% | 0 | 0 | — |
case-05 | fail→fail | 26,776 | 28,081 | +5% | 1 | 1 | 0% | 3,737 | 1,283 | -66% | 0 | 0 | — |
case-06 | fail→fail | 41,728 | 58,416 | +40% | 1 | 1 | 0% | 6,381 | 1,130 | -82% | 0 | 0 | — |
case-07 | fail→fail | 26,563 | 24,908 | -6% | 1 | 1 | 0% | 3,493 | 1,095 | -69% | 0 | 0 | — |
case-08 | fail→fail | 26,093 | 18,647 | -29% | 1 | 1 | 0% | 3,345 | 466 | -86% | 0 | 0 | — |
case-09 | fail→fail | 48,207 | 17,980 | -63% | 1 | 1 | 0% | 8,219 | 474 | -94% | 0 | 0 | — |
case-10 | fail→fail | 24,869 | 52,524 | +111% | 1 | 1 | 0% | 3,327 | 8,371 | +152% | 0 | 0 | — |
case-11 | fail→fail | 51,605 | 56,166 | +9% | 1 | 1 | 0% | 8,225 | 7,158 | -13% | 0 | 0 | — |
case-14 | fail→fail | 29,370 | 28,473 | -3% | 1 | 1 | 0% | 3,799 | 2,202 | -42% | 0 | 0 | — |
case-15 | fail→fail | 37,854 | 46,773 | +24% | 1 | 1 | 0% | 5,942 | 8,373 | +41% | 0 | 0 | — |
case-16 | fail→fail | 31,855 | 50,987 | +60% | 1 | 1 | 0% | 4,677 | 8,370 | +79% | 0 | 0 | — |
case-17 | fail→fail | 41,331 | 52,576 | +27% | 1 | 1 | 0% | 5,872 | 8,381 | +43% | 0 | 0 | — |
case-18 | pass→pass | 16,618 | 8,323 | -50% | 1 | 1 | 0% | 1,699 | 610 | -64% | 0 | 0 | — |
case-19 | fail→pass | 25,703 | 7,885 | -69% | 1 | 1 | 0% | 3,144 | 616 | -80% | 0 | 0 | — |
case-20 | pass→pass | 16,051 | 14,245 | -11% | 1 | 1 | 0% | 1,904 | 1,718 | -10% | 0 | 0 | — |
case-21 | pass→fail | 13,607 | 8,747 | -36% | 1 | 1 | 0% | 1,528 | 724 | -53% | 0 | 0 | — |
case-22 | pass→pass | 11,633 | 13,012 | +12% | 1 | 1 | 0% | 1,146 | 1,636 | +43% | 0 | 0 | — |
case-23 | pass→pass | 13,400 | 15,450 | +15% | 1 | 1 | 0% | 1,541 | 2,086 | +35% | 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 14 counted toward the lift figure. The other 9 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 0 percentage points is the difference between those two pass rates over the 14 comparable cases. 1 case got worse with the skill loaded, and it is 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.