Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate solutions targeting specific coverage gaps — detect gaps, generate failure-driven solutions, and design factor-level experiments.
.claude/skills/yogsoth-ai-gap-driven-generation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 108% | 0% |
Detect coverage gaps in the solution space, then generate targeted solutions using failure analysis and factorial design.
Run coverage-gap-detection SOP to identify uncovered regions in the solution space. Output: gap list with priority ranking based on impact and feasibility.
Run failure-driven-generation SOP to generate solutions that specifically target identified failure modes or gaps. Output: one or more targeted solutions per gap.
Run factor-level-design SOP to identify key factors in the gap space, define levels, and design an experiment matrix for systematic exploration. Output: factorial design ready for implementation.
| Metric | Floor | |--------|-------| | Coverage gaps identified | ≥3 | | Gaps with priority ranking | 100% | | Solutions generated per gap | ≥1 | | Total targeted solutions | ≥3 | | Factor-level matrix produced | yes |
| SOP | Role | |-----|------| | coverage-gap-detection | Stage 1 — detect uncovered regions | | failure-driven-generation | Stage 2 — generate targeted solutions | | factor-level-design | Stage 3 — factorial experiment design |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | coverage-gap-detection | Detect uncovered regions in the solution space, producing a prioritized gap list. | | creative-ideation-factor-level-design | Identify factors and their levels for a problem, then design an experiment matrix for systematic exploration. | | creative-ideation-failure-mode-cataloging | Systematically catalog all failure modes in a domain or method, producing a classified failure taxonomy. | | failure-driven-generation | Generate targeted solutions for each identified failure mode, ensuring every failure has at least one proposed mitigation. |
<!-- 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-05 | fail→pass | 32,985 | 26,051 | -21% | 1 | 1 | 0% | 6,184 | 5,180 | -16% | 0 | 0 | — |
case-06 | fail→pass | 31,418 | 23,138 | -26% | 1 | 1 | 0% | 5,096 | 4,477 | -12% | 0 | 0 | — |
case-07 | fail→pass | 23,214 | 22,375 | -4% | 1 | 1 | 0% | 3,881 | 4,183 | +8% | 0 | 0 | — |
case-08 | fail→pass | 19,943 | 24,223 | +21% | 1 | 1 | 0% | 2,676 | 4,374 | +63% | 0 | 0 | — |
case-09 | fail→pass | 21,021 | 30,042 | +43% | 1 | 1 | 0% | 2,695 | 5,596 | +108% | 0 | 0 | — |
case-01 | fail→pass | 43,940 | 25,792 | -41% | 1 | 1 | 0% | 2,119 | 4,612 | +118% | 0 | 0 | — |
case-02 | fail→pass | 23,469 | 17,698 | -25% | 1 | 1 | 0% | 3,678 | 3,364 | -9% | 0 | 0 | — |
case-03 | pass→pass | 18,082 | 27,656 | +53% | 1 | 1 | 0% | 2,852 | 4,872 | +71% | 0 | 0 | — |
case-04 | fail→pass | 31,259 | 36,327 | +16% | 1 | 1 | 0% | 5,254 | 6,660 | +27% | 0 | 0 | — |
case-10 | fail→pass | 34,114 | 27,936 | -18% | 1 | 1 | 0% | 6,178 | 5,650 | -9% | 0 | 0 | — |
case-11 | fail→pass | 28,329 | 28,982 | +2% | 1 | 1 | 0% | 4,631 | 5,269 | +14% | 0 | 0 | — |
case-12 | fail→pass | 33,967 | 31,139 | -8% | 1 | 1 | 0% | 5,794 | 6,126 | +6% | 0 | 0 | — |
case-13 | fail→pass | 35,914 | 35,876 | -0% | 1 | 1 | 0% | 6,186 | 6,644 | +7% | 0 | 0 | — |
case-14 | fail→pass | 33,319 | 30,457 | -9% | 1 | 1 | 0% | 6,186 | 6,222 | +1% | 0 | 0 | — |
case-15 | fail→fail | 37,325 | 38,308 | +3% | 1 | 1 | 0% | 6,189 | 6,647 | +7% | 0 | 0 | — |
case-16 | fail→pass | 38,883 | 35,524 | -9% | 1 | 1 | 0% | 6,186 | 6,270 | +1% | 0 | 0 | — |
case-17 | fail→pass | 39,178 | 33,112 | -15% | 1 | 1 | 0% | 6,188 | 6,204 | +0% | 0 | 0 | — |
case-18 | fail→pass | 35,382 | 35,698 | +1% | 1 | 1 | 0% | 6,108 | 6,645 | +9% | 0 | 0 | — |
case-19 | fail→pass | 32,060 | 29,823 | -7% | 1 | 1 | 0% | 4,989 | 5,524 | +11% | 0 | 0 | — |
case-20 | pass→fail | 16,125 | 31,716 | +97% | 1 | 1 | 0% | 2,730 | 6,036 | +121% | 0 | 0 | — |
case-21 | pass→fail | 17,838 | 23,418 | +31% | 1 | 1 | 0% | 3,025 | 4,596 | +52% | 0 | 0 | — |
case-22 | pass→pass | 14,726 | 21,417 | +45% | 1 | 1 | 0% | 3,233 | 4,938 | +53% | 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 +68 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 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.