Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Multi-dimensional evaluation and tiered filtering of generated ideas. Orchestrates novelty assessment → feasibility check → ranking → selection.
.claude/skills/yogsoth-ai-evaluation-filtering/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-16 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 68% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -13% | 0% |
Multi-dimensional evaluation and tiered filtering of generated ideas.
Score all ideas on novelty dimensions using novelty-scoring SOP. Assign tier: BREAKTHROUGH / HIGH / MODERATE / INCREMENTAL.
For HIGH and BREAKTHROUGH tier ideas, assess initial feasibility:
This is a signal, not a deep validation — that's the convergence repo's job.
Rank ideas by composite score: novelty (0.6) × feasibility signal (0.2) × completeness (0.2).
Select top ideas for output. All BREAKTHROUGH ideas pass regardless of feasibility. HIGH ideas pass if feasibility ≥ MEDIUM. MODERATE ideas pass only if they fill a coverage gap.
| Metric | Floor | |--------|-------| | Ideas evaluated | all generated ideas | | Ideas scored on all dimensions | 100% | | Tier distribution reported | yes | | Top ideas selected | ≥5 (or all BREAKTHROUGH + HIGH) |
| SOP | Role | |-----|------| | novelty-scoring | Stage 1 — multi-dimensional novelty assessment | | constraint-injection | Stage 2 — test feasibility under constraints | | idea-synthesis | Stage 3 — refine top ideas into complete descriptions | | saturation-detection | Pre — determine if enough ideas exist to evaluate |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-16 | fail→pass | 13,138 | 4,677 | -64% | 1 | 1 | 0% | 1,935 | 1,179 | -39% | 0 | 0 | — |
case-01 | fail→pass | 6,525 | 5,255 | -19% | 1 | 1 | 0% | 1,187 | 1,236 | +4% | 0 | 0 | — |
case-02 | fail→pass | 31,101 | 21,634 | -30% | 1 | 1 | 0% | 5,748 | 4,183 | -27% | 0 | 0 | — |
case-03 | fail→pass | 19,899 | 26,492 | +33% | 1 | 1 | 0% | 3,321 | 5,583 | +68% | 0 | 0 | — |
case-04 | fail→pass | 10,638 | 6,012 | -43% | 1 | 1 | 0% | 1,664 | 1,448 | -13% | 0 | 0 | — |
case-05 | pass→pass | 9,016 | 2,157 | -76% | 1 | 1 | 0% | 1,344 | 744 | -45% | 0 | 0 | — |
case-06 | fail→fail | 11,071 | 5,424 | -51% | 1 | 1 | 0% | 1,772 | 1,142 | -36% | 0 | 0 | — |
case-07 | pass→fail | 13,978 | 6,382 | -54% | 1 | 1 | 0% | 2,415 | 1,486 | -38% | 0 | 0 | — |
case-08 | pass→pass | 15,355 | 4,869 | -68% | 1 | 1 | 0% | 2,392 | 1,176 | -51% | 0 | 0 | — |
case-09 | fail→pass | 11,103 | 4,940 | -56% | 1 | 1 | 0% | 1,592 | 1,091 | -31% | 0 | 0 | — |
case-10 | fail→fail | 3,808 | 5,988 | +57% | 1 | 1 | 0% | 589 | 1,335 | +127% | 0 | 0 | — |
case-11 | fail→fail | 13,966 | 5,996 | -57% | 1 | 1 | 0% | 2,064 | 1,243 | -40% | 0 | 0 | — |
case-12 | fail→pass | 11,099 | 1,965 | -82% | 1 | 1 | 0% | 1,704 | 667 | -61% | 0 | 0 | — |
case-13 | pass→fail | 9,058 | 2,064 | -77% | 1 | 1 | 0% | 1,406 | 688 | -51% | 0 | 0 | — |
case-14 | fail→pass | 10,077 | 1,505 | -85% | 1 | 1 | 0% | 1,582 | 592 | -63% | 0 | 0 | — |
case-15 | pass→pass | 8,849 | 1,898 | -79% | 1 | 1 | 0% | 1,372 | 652 | -52% | 0 | 0 | — |
case-22 | pass→fail | 25,051 | 32,172 | +28% | 1 | 1 | 0% | 4,348 | 6,431 | +48% | 0 | 0 | — |
case-17 | fail→pass | 12,919 | 4,370 | -66% | 1 | 1 | 0% | 2,015 | 1,023 | -49% | 0 | 0 | — |
case-18 | pass→pass | 10,624 | 4,092 | -61% | 1 | 1 | 0% | 1,677 | 1,039 | -38% | 0 | 0 | — |
case-19 | pass→pass | 13,161 | 2,754 | -79% | 1 | 1 | 0% | 1,856 | 754 | -59% | 0 | 0 | — |
case-20 | pass→pass | 18,852 | 12,313 | -35% | 1 | 1 | 0% | 2,837 | 2,247 | -21% | 0 | 0 | — |
case-21 | pass→pass | 19,525 | 21,526 | +10% | 1 | 1 | 0% | 3,067 | 3,964 | +29% | 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 +27 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 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.