Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write clear design rationale connecting decisions to user needs, business goals, and principles.
.claude/skills/owl-listener-design-rationale/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 7% | 0% |
You are an expert in articulating the reasoning behind design decisions.
You write clear design rationale that connects decisions to evidence, principles, and goals.
What design decision was made? Be specific about what was chosen.
What problem or need prompted this decision? What constraints exist?
What alternatives were explored? Brief description of each.
What informed the decision? User research, data, best practices, competitive analysis, usability testing.
Why this option over the alternatives? Connect to user needs, business goals, design principles, and technical feasibility.
What are the known compromises? What was deprioritized and why?
How will you know if this decision was right? What metrics or feedback will confirm?
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,250 | 13,774 | -35% | 1 | 1 | 0% | 3,557 | 2,799 | -21% | 0 | 0 | — |
case-02 | fail→pass | 17,960 | 13,030 | -27% | 1 | 1 | 0% | 2,902 | 2,533 | -13% | 0 | 0 | — |
case-03 | fail→pass | 16,518 | 13,726 | -17% | 1 | 1 | 0% | 2,583 | 2,582 | -0% | 0 | 0 | — |
case-04 | fail→pass | 17,682 | 13,936 | -21% | 1 | 1 | 0% | 2,689 | 2,590 | -4% | 0 | 0 | — |
case-05 | pass→pass | 13,798 | 13,073 | -5% | 1 | 1 | 0% | 2,439 | 2,344 | -4% | 0 | 0 | — |
case-06 | pass→pass | 14,395 | 14,100 | -2% | 1 | 1 | 0% | 2,490 | 2,864 | +15% | 0 | 0 | — |
case-07 | fail→pass | 14,453 | 14,375 | -1% | 1 | 1 | 0% | 2,249 | 2,406 | +7% | 0 | 0 | — |
case-08 | fail→pass | 16,792 | 12,538 | -25% | 1 | 1 | 0% | 2,942 | 2,859 | -3% | 0 | 0 | — |
case-09 | pass→fail | 18,935 | 14,088 | -26% | 1 | 1 | 0% | 2,908 | 2,569 | -12% | 0 | 0 | — |
case-10 | pass→pass | 14,882 | 12,729 | -14% | 1 | 1 | 0% | 2,335 | 2,384 | +2% | 0 | 0 | — |
case-11 | fail→fail | 14,604 | 11,458 | -22% | 1 | 1 | 0% | 2,401 | 2,274 | -5% | 0 | 0 | — |
case-12 | fail→pass | 14,886 | 10,797 | -27% | 1 | 1 | 0% | 2,288 | 2,045 | -11% | 0 | 0 | — |
case-13 | fail→pass | 12,437 | 12,315 | -1% | 1 | 1 | 0% | 2,001 | 2,371 | +18% | 0 | 0 | — |
case-14 | fail→pass | 15,050 | 12,656 | -16% | 1 | 1 | 0% | 2,522 | 2,330 | -8% | 0 | 0 | — |
case-15 | fail→pass | 13,326 | 10,975 | -18% | 1 | 1 | 0% | 2,035 | 2,106 | +3% | 0 | 0 | — |
case-16 | fail→fail | 13,314 | 15,100 | +13% | 1 | 1 | 0% | 2,089 | 2,579 | +23% | 0 | 0 | — |
case-17 | fail→pass | 15,201 | 14,308 | -6% | 1 | 1 | 0% | 2,347 | 2,765 | +18% | 0 | 0 | — |
case-18 | fail→pass | 14,788 | 14,562 | -2% | 1 | 1 | 0% | 2,506 | 2,760 | +10% | 0 | 0 | — |
case-19 | pass→pass | 15,429 | 12,304 | -20% | 1 | 1 | 0% | 2,423 | 2,412 | -0% | 0 | 0 | — |
case-20 | pass→fail | 6,983 | 13,254 | +90% | 1 | 1 | 0% | 1,358 | 2,640 | +94% | 0 | 0 | — |
case-21 | fail→fail | 12,718 | 12,408 | -2% | 1 | 1 | 0% | 2,316 | 2,475 | +7% | 0 | 0 | — |
case-22 | pass→pass | 9,437 | 9,271 | -2% | 1 | 1 | 0% | 1,591 | 1,965 | +24% | 0 | 0 | — |
case-23 | pass→fail | 14,515 | 13,900 | -4% | 1 | 1 | 0% | 2,852 | 2,916 | +2% | 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. The headline lift of +39 percentage points is the difference between those two pass rates over the 23 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.