Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build performance-over-time progress curves with inflection detection
.claude/skills/yogsoth-ai-progress-curve-construction/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 120% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 43% | 0% |
Transform historical performance data into temporal progress curves that reveal the rate of improvement, paradigm shifts, and saturation trends. Enables quantitative headroom analysis and identifies where the field needs breakthroughs vs. incremental refinement.
Organize all historical scores into chronological sequences:
Yield: Chronological score sequences per benchmark.
Fit parametric models to the SOTA frontier:
Yield: Fitted curves with parameters and confidence bands.
Identify moments where progress rate changed significantly:
Yield: Annotated inflection points with causal attribution.
Quantify remaining improvement potential:
Yield: Headroom estimates with confidence intervals.
| Metric | Floor | |--------|-------| | Progress curves constructed | 2 | | Years of history covered | 3 | | Inflection points identified | 1 | | Headroom estimates produced | 2 |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | baseline-synthesis | Produce final structured baseline report integrating all analysis results | | headroom-estimation | Estimate theoretical/practical ceiling vs current SOTA gap | | progress-curve-fitting | Construct performance-over-time visualization data |
<!-- 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-19 | fail→pass | 19,782 | 14,054 | -29% | 1 | 1 | 0% | 2,212 | 2,247 | +2% | 0 | 0 | — |
case-01 | fail→fail | 42,863 | 44,316 | +3% | 1 | 1 | 0% | 8,264 | 8,123 | -2% | 0 | 0 | — |
case-02 | fail→pass | 23,817 | 37,828 | +59% | 1 | 1 | 0% | 3,412 | 7,502 | +120% | 0 | 0 | — |
case-03 | fail→fail | 19,652 | 29,521 | +50% | 1 | 1 | 0% | 3,482 | 4,855 | +39% | 0 | 0 | — |
case-04 | pass→pass | 24,212 | 26,610 | +10% | 1 | 1 | 0% | 3,153 | 4,496 | +43% | 0 | 0 | — |
case-05 | pass→pass | 22,859 | 25,405 | +11% | 1 | 1 | 0% | 3,056 | 4,300 | +41% | 0 | 0 | — |
case-06 | fail→fail | 49,792 | 20,988 | -58% | 1 | 1 | 0% | 3,219 | 3,139 | -2% | 0 | 0 | — |
case-07 | fail→pass | 16,307 | 14,668 | -10% | 1 | 1 | 0% | 1,826 | 1,876 | +3% | 0 | 0 | — |
case-08 | pass→pass | 19,698 | 19,757 | +0% | 1 | 1 | 0% | 3,040 | 2,949 | -3% | 0 | 0 | — |
case-09 | pass→pass | 20,360 | 15,936 | -22% | 1 | 1 | 0% | 2,442 | 2,205 | -10% | 0 | 0 | — |
case-10 | pass→pass | 27,958 | 22,321 | -20% | 1 | 1 | 0% | 3,497 | 3,417 | -2% | 0 | 0 | — |
case-11 | fail→fail | 13,086 | 8,567 | -35% | 1 | 1 | 0% | 1,402 | 1,170 | -17% | 0 | 0 | — |
case-12 | fail→fail | 19,722 | 3,746 | -81% | 1 | 1 | 0% | 2,165 | 1,173 | -46% | 0 | 0 | — |
case-13 | pass→pass | 34,453 | 7,642 | -78% | 1 | 1 | 0% | 1,940 | 896 | -54% | 0 | 0 | — |
case-18 | pass→pass | 30,890 | 31,897 | +3% | 1 | 1 | 0% | 3,748 | 4,658 | +24% | 0 | 0 | — |
case-14 | pass→pass | 36,684 | 8,821 | -76% | 1 | 1 | 0% | 2,459 | 1,193 | -51% | 0 | 0 | — |
case-15 | fail→pass | 18,686 | 16,821 | -10% | 1 | 1 | 0% | 2,347 | 2,545 | +8% | 0 | 0 | — |
case-16 | fail→fail | 19,994 | 31,467 | +57% | 1 | 1 | 0% | 2,264 | 2,493 | +10% | 0 | 0 | — |
case-17 | fail→fail | 22,294 | 21,863 | -2% | 1 | 1 | 0% | 2,668 | 3,200 | +20% | 0 | 0 | — |
case-20 | fail→fail | 14,811 | 39,873 | +169% | 1 | 1 | 0% | 1,825 | 6,948 | +281% | 0 | 0 | — |
case-21 | fail→fail | 20,320 | 22,174 | +9% | 1 | 1 | 0% | 3,514 | 3,417 | -3% | 0 | 0 | — |
case-22 | fail→fail | 13,191 | 26,911 | +104% | 1 | 1 | 0% | 2,222 | 5,291 | +138% | 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 +18 percentage points is the difference between those two pass rates over the 22 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.