Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Critique a screen's visual hierarchy — entry point, eye flow, weight distribution, and emphasis.
.claude/skills/owl-listener-critique-visual-hierarchy/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 94% | 20 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 10% | 0% |
You are an expert in visual hierarchy and screen-level design critique.
You analyse a screen to identify whether hierarchy is clear, intentional, and aligned with user goals. You flag problems and suggest targeted fixes.
The first element that captures the eye. Evaluate whether it is the most important thing on screen.
The path a user's eye travels after landing. Evaluate whether the path is deliberate and efficient.
The relative visual importance of each element. Evaluate whether weight is distributed purposefully.
Specific elements that demand extra attention. Evaluate whether emphasis is earned and singular.
For each dimension — Entry Point, Eye Flow, Weight, Emphasis — provide:
Rate each dimension: pass / minor issue / major issue.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,300 | 15,638 | +2% | 1 | 1 | 0% | 2,664 | 3,144 | +18% | 0 | 0 | — |
case-02 | fail→pass | 19,930 | 16,560 | -17% | 1 | 1 | 0% | 3,274 | 3,097 | -5% | 0 | 0 | — |
case-03 | fail→pass | 12,478 | 13,633 | +9% | 1 | 1 | 0% | 2,263 | 2,875 | +27% | 0 | 0 | — |
case-04 | pass→pass | 8,459 | 6,889 | -19% | 1 | 1 | 0% | 1,649 | 1,673 | +1% | 0 | 0 | — |
case-05 | fail→pass | 12,482 | 8,893 | -29% | 1 | 1 | 0% | 2,512 | 2,117 | -16% | 0 | 0 | — |
case-06 | pass→pass | 13,426 | 9,457 | -30% | 1 | 1 | 0% | 2,121 | 2,004 | -6% | 0 | 0 | — |
case-07 | pass→pass | 8,802 | 8,380 | -5% | 1 | 1 | 0% | 1,991 | 2,185 | +10% | 0 | 0 | — |
case-08 | pass→pass | 12,685 | 13,923 | +10% | 1 | 1 | 0% | 2,244 | 2,996 | +34% | 0 | 0 | — |
case-09 | pass→pass | 10,956 | 8,885 | -19% | 1 | 1 | 0% | 2,534 | 2,728 | +8% | 0 | 0 | — |
case-10 | pass→pass | 12,503 | 4,591 | -63% | 1 | 1 | 0% | 2,053 | 1,290 | -37% | 0 | 0 | — |
case-11 | fail→pass | 11,712 | 11,171 | -5% | 1 | 1 | 0% | 1,915 | 2,102 | +10% | 0 | 0 | — |
case-12 | fail→pass | 13,622 | 8,297 | -39% | 1 | 1 | 0% | 2,239 | 1,774 | -21% | 0 | 0 | — |
case-13 | pass→pass | 11,820 | 5,070 | -57% | 1 | 1 | 0% | 2,094 | 1,319 | -37% | 0 | 0 | — |
case-14 | fail→pass | 14,003 | 10,869 | -22% | 1 | 1 | 0% | 2,310 | 2,317 | +0% | 0 | 0 | — |
case-15 | pass→pass | 10,489 | 10,090 | -4% | 1 | 1 | 0% | 1,603 | 2,147 | +34% | 0 | 0 | — |
case-16 | fail→pass | 14,745 | 16,507 | +12% | 1 | 1 | 0% | 2,452 | 3,241 | +32% | 0 | 0 | — |
case-17 | fail→pass | 8,720 | 10,490 | +20% | 1 | 1 | 0% | 1,582 | 2,211 | +40% | 0 | 0 | — |
case-18 | fail→fail | 13,513 | 4,871 | -64% | 1 | 1 | 0% | 2,098 | 1,319 | -37% | 0 | 0 | — |
case-19 | fail→fail | 13,226 | 7,210 | -45% | 1 | 1 | 0% | 2,212 | 1,699 | -23% | 0 | 0 | — |
case-20 | fail→fail | 12,771 | 6,075 | -52% | 1 | 1 | 0% | 2,043 | 1,528 | -25% | 0 | 0 | — |
case-21 | pass→pass | 4,223 | 8,337 | +97% | 1 | 1 | 0% | 779 | 1,924 | +147% | 0 | 0 | — |
case-22 | fail→pass | 12,024 | 8,479 | -29% | 1 | 1 | 0% | 1,811 | 1,887 | +4% | 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 +45 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.