Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Strategy: Select an RQ framework (PICO/SPIDER/SPICE/ECLIPSE) and apply it systematically
.claude/skills/yogsoth-ai-framework-guided-formulation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 98% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 31% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 8% | 0% |
Select the most suitable RQ framework and apply it systematically — when the research type is clear, use a standard framework to structure the question.
Core logic: Different research types have different standards for a "good question." A framework is a checklist, distilled by predecessors, of "what components a good question should contain." Pick the right framework → fill in each component → naturally arrive at a structured question.
| Research type | Recommended framework | Core components | |----------|---------|---------| | Quantitative/intervention research | PICO/PICOTS | Population, Intervention, Comparison, Outcome (+Time, Setting) | | Qualitative research | SPIDER | Sample, Phenomenon of Interest, Design, Evaluation, Research type | | Evaluation research | SPICE | Setting, Perspective, Intervention, Comparison, Evaluation | | Mixed methods | ECLIPSE | Expectation, Client group, Location, Impact, Professionals, Service |
| Tier | Framework candidates | Framework application | FINER check | Output | |------|---------|---------|-----------|------| | S | ≥2 frameworks compared | Selected framework, all components filled | All 5 items pass | ≥1 RQ | | M | ≥3 frameworks compared | Selected framework, all components + alternative framework comparison | 5 items + success criteria | ≥2 RQ | | L | ≥4 frameworks compared | Multiple frameworks applied in parallel + optimal selection | 5 items + criteria + sensitivity | ≥3 RQ |
After the Strategy completes, context-checkpoint must be called, recording:
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | framework-selection-and-application | Tactic: Select the most suitable RQ framework and apply it systematically | | question-refinement-loop | Tactic: iteratively refine a research question until it passes all 5 FINER criteria |
<!-- 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→pass | 17,024 | 11,847 | -30% | 1 | 1 | 0% | 2,871 | 2,817 | -2% | 0 | 0 | — |
case-13 | pass→pass | 16,462 | 21,148 | +28% | 1 | 1 | 0% | 2,711 | 4,449 | +64% | 0 | 0 | — |
case-19 | fail→pass | 9,317 | 14,055 | +51% | 1 | 1 | 0% | 1,564 | 3,096 | +98% | 0 | 0 | — |
case-01 | fail→pass | 17,178 | 22,701 | +32% | 1 | 1 | 0% | 2,863 | 3,072 | +7% | 0 | 0 | — |
case-02 | fail→pass | 16,013 | 19,576 | +22% | 1 | 1 | 0% | 2,640 | 3,465 | +31% | 0 | 0 | — |
case-03 | fail→pass | 13,119 | 9,865 | -25% | 1 | 1 | 0% | 2,201 | 2,380 | +8% | 0 | 0 | — |
case-04 | pass→fail | 18,117 | 20,978 | +16% | 1 | 1 | 0% | 3,483 | 4,952 | +42% | 0 | 0 | — |
case-05 | pass→fail | 10,900 | 23,816 | +118% | 1 | 1 | 0% | 2,092 | 5,212 | +149% | 0 | 0 | — |
case-06 | pass→pass | 13,677 | 16,362 | +20% | 1 | 1 | 0% | 2,397 | 3,540 | +48% | 0 | 0 | — |
case-07 | fail→fail | 2,646 | 3,573 | +35% | 1 | 1 | 0% | 495 | 1,280 | +159% | 0 | 0 | — |
case-08 | fail→pass | 2,503 | 8,983 | +259% | 1 | 1 | 0% | 426 | 2,193 | +415% | 0 | 0 | — |
case-09 | fail→fail | 2,931 | 3,733 | +27% | 1 | 1 | 0% | 548 | 1,302 | +138% | 0 | 0 | — |
case-10 | fail→pass | 4,546 | 10,371 | +128% | 1 | 1 | 0% | 686 | 2,642 | +285% | 0 | 0 | — |
case-11 | fail→pass | 20,041 | 12,632 | -37% | 1 | 1 | 0% | 3,671 | 2,949 | -20% | 0 | 0 | — |
case-14 | fail→pass | 35,689 | 30,870 | -14% | 1 | 1 | 0% | 6,182 | 6,048 | -2% | 0 | 0 | — |
case-15 | fail→pass | 9,860 | 14,324 | +45% | 1 | 1 | 0% | 1,558 | 2,916 | +87% | 0 | 0 | — |
case-16 | pass→pass | 15,073 | 14,301 | -5% | 1 | 1 | 0% | 2,316 | 3,292 | +42% | 0 | 0 | — |
case-17 | fail→pass | 23,049 | 42,411 | +84% | 1 | 1 | 0% | 3,690 | 4,649 | +26% | 0 | 0 | — |
case-18 | fail→pass | 13,888 | 11,893 | -14% | 1 | 1 | 0% | 2,062 | 2,507 | +22% | 0 | 0 | — |
case-20 | fail→fail | 9,809 | 11,858 | +21% | 1 | 1 | 0% | 1,903 | 2,780 | +46% | 0 | 0 | — |
case-21 | fail→pass | 7,751 | 13,674 | +76% | 1 | 1 | 0% | 1,227 | 2,991 | +144% | 0 | 0 | — |
case-22 | fail→pass | 16,790 | 16,353 | -3% | 1 | 1 | 0% | 2,646 | 3,513 | +33% | 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 +55 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.