Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Apply Kuhn's paradigm theory to analyze scientific progress through the cycle of normal science, anomalies, crisis, and revolution. Use this skill when the user needs to understand why a field resists change, trace paradigm shifts in a discipline, analyze incommensurability between competing frameworks, or when they ask 'why do scientists ignore contradictory evidence', 'how do scientific revolutions happen', or 'why can't proponents of different paradigms agree'.
.claude/skills/asgard-ai-platform-grad-paradigms/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-23 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-03 | ✓→✗ | ▼ Worse | 54% | 0% |
| case-12 | ✓→✓ | = Same ✓ | 45% | 0% |
Kuhn's paradigm theory explains scientific progress not as linear accumulation but as a cyclical process: normal science operates within a paradigm until anomalies accumulate, triggering crisis and eventually a revolutionary paradigm shift. The new paradigm is incommensurable with the old — they literally see different worlds.
IRON LAW: Scientists working within a paradigm do NOT test the paradigm —
they solve puzzles defined by it. Paradigm change is a social-political
process, not a purely rational one.Key assumptions:
Define the dominant paradigm: shared exemplars, accepted methods, ontological commitments, and the community that holds them.
Identify puzzle-solving within the paradigm — what questions are considered legitimate, what methods are standard, what counts as a solution.
Document anomalies (persistent puzzles the paradigm cannot solve), and assess whether they have accumulated to crisis level — marked by proliferation of ad hoc modifications and questioning of fundamentals.
Determine whether a rival paradigm has emerged, evaluate incommensurability with the old paradigm, and trace the social process of conversion (generational replacement, institutional power shifts).
markdown## Paradigm Analysis: [Context] ### Dominant Paradigm - Core exemplars: [foundational achievements that define the paradigm] - Disciplinary matrix: [shared values, methods, symbolic generalizations] - Community: [who subscribes to this paradigm] ### Normal Science Phase - Legitimate puzzles: [questions the paradigm defines as worth solving] - Standard methods: [accepted approaches] - Anomalies identified: [persistent unsolved puzzles] ### Crisis Assessment - Severity: [pre-crisis / emerging crisis / full crisis] - Ad hoc modifications: [patches to save the paradigm] - Competing candidates: [alternative paradigms emerging] ### Paradigm Shift Evaluation - Rival paradigm: [description if one exists] - Incommensurability points: [where old and new paradigms talk past each other] - Conversion dynamics: [generational, institutional, evidential factors] ### Implications 1. [Current phase of the field] 2. [Likelihood and direction of potential shift]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | pass→pass | 16,924 | 19,138 | +13% | 1 | 1 | 0% | 2,410 | 3,490 | +45% | 0 | 0 | — |
case-01 | fail→fail | 49,377 | 44,989 | -9% | 1 | 1 | 0% | 7,497 | 5,364 | -28% | 0 | 0 | — |
case-02 | fail→fail | 46,406 | 27,056 | -42% | 1 | 1 | 0% | 7,609 | 4,872 | -36% | 0 | 0 | — |
case-03 | pass→fail | 23,313 | 26,729 | +15% | 1 | 1 | 0% | 3,496 | 5,387 | +54% | 0 | 0 | — |
case-04 | pass→pass | 18,476 | 26,820 | +45% | 1 | 1 | 0% | 2,743 | 4,564 | +66% | 0 | 0 | — |
case-05 | fail→fail | 53,637 | 60,649 | +13% | 1 | 1 | 0% | 1,380 | 3,920 | +184% | 0 | 0 | — |
case-06 | pass→pass | 15,719 | 13,844 | -12% | 1 | 1 | 0% | 2,149 | 3,090 | +44% | 0 | 0 | — |
case-07 | pass→pass | 13,635 | 12,677 | -7% | 1 | 1 | 0% | 2,117 | 2,951 | +39% | 0 | 0 | — |
case-08 | pass→pass | 15,165 | 16,516 | +9% | 1 | 1 | 0% | 2,118 | 3,407 | +61% | 0 | 0 | — |
case-09 | pass→pass | 17,253 | 18,432 | +7% | 1 | 1 | 0% | 2,668 | 3,755 | +41% | 0 | 0 | — |
case-10 | pass→pass | 15,702 | 12,849 | -18% | 1 | 1 | 0% | 2,468 | 2,897 | +17% | 0 | 0 | — |
case-11 | pass→pass | 7,705 | 10,791 | +40% | 1 | 1 | 0% | 1,284 | 2,732 | +113% | 0 | 0 | — |
case-13 | pass→pass | 14,454 | 14,893 | +3% | 1 | 1 | 0% | 2,141 | 2,990 | +40% | 0 | 0 | — |
case-14 | fail→pass | 29,519 | 17,636 | -40% | 1 | 1 | 0% | 4,931 | 3,888 | -21% | 0 | 0 | — |
case-15 | fail→fail | 37,434 | 20,436 | -45% | 1 | 1 | 0% | 5,985 | 4,087 | -32% | 0 | 0 | — |
case-16 | fail→pass | 31,626 | 13,841 | -56% | 1 | 1 | 0% | 4,967 | 3,236 | -35% | 0 | 0 | — |
case-17 | pass→pass | 13,212 | 16,210 | +23% | 1 | 1 | 0% | 2,056 | 3,089 | +50% | 0 | 0 | — |
case-18 | pass→pass | 11,716 | 10,516 | -10% | 1 | 1 | 0% | 1,620 | 2,620 | +62% | 0 | 0 | — |
case-19 | pass→pass | 18,300 | 14,882 | -19% | 1 | 1 | 0% | 2,284 | 3,018 | +32% | 0 | 0 | — |
case-20 | pass→pass | 14,019 | 15,275 | +9% | 1 | 1 | 0% | 1,869 | 3,179 | +70% | 0 | 0 | — |
case-21 | pass→pass | 12,996 | 13,712 | +6% | 1 | 1 | 0% | 2,139 | 2,780 | +30% | 0 | 0 | — |
case-22 | pass→pass | 13,729 | 17,189 | +25% | 1 | 1 | 0% | 2,106 | 3,589 | +70% | 0 | 0 | — |
case-23 | fail→pass | 50,059 | 20,051 | -60% | 1 | 1 | 0% | 8,227 | 4,210 | -49% | 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 22 counted toward the lift figure. The other 1 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 +9 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.