Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Visual regression testing — capture before/after screenshots of UI changes and compare. Use when a UI change might have unintended visual impact across components.
.claude/skills/evolution-foundation-dev-visual-verdict/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 369% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -23% | 0% |
Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.
Visual regression testing. Capture before/after screenshots of UI components, compare them, and flag unintended visual impact.
@canvas-designer instead)@bolt-executor make the modificationSaved to workspace/development/verifications/[C]visual-{component}-{date}.md:
markdown## Visual Verdict — {component} ### Scope - Components checked: [list] ### Comparison | Component | Before | After | Diff | Verdict | |---|---|---|---|---| | Header |  |  | 0% | IDENTICAL | | Footer |  |  | 12% | UNEXPECTED | ### Verdict {IDENTICAL | EXPECTED CHANGES | UNEXPECTED REGRESSION} ### Unexpected Regressions - Footer color shifted from #1A1A1A to #2A2A2A — investigate cascade
@canvas-designer (when regression is intentional)@hawk-debugger (when regression is unintended)@oath-verifier (final visual verification)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 3,236 | 8,461 | +161% | 1 | 1 | 0% | 413 | 1,937 | +369% | 0 | 0 | — |
case-02 | fail→fail | 4,563 | 7,884 | +73% | 1 | 1 | 0% | 260 | 665 | +156% | 0 | 0 | — |
case-03 | fail→fail | 17,476 | 8,331 | -52% | 1 | 1 | 0% | 2,960 | 1,454 | -51% | 0 | 0 | — |
case-04 | pass→pass | 13,504 | 5,135 | -62% | 1 | 1 | 0% | 1,997 | 1,291 | -35% | 0 | 0 | — |
case-05 | fail→pass | 11,834 | 4,389 | -63% | 1 | 1 | 0% | 1,860 | 1,159 | -38% | 0 | 0 | — |
case-06 | pass→pass | 8,965 | 3,867 | -57% | 1 | 1 | 0% | 1,402 | 1,067 | -24% | 0 | 0 | — |
case-07 | fail→pass | 14,832 | 6,972 | -53% | 1 | 1 | 0% | 2,322 | 1,676 | -28% | 0 | 0 | — |
case-08 | fail→pass | 10,750 | 2,830 | -74% | 1 | 1 | 0% | 1,590 | 909 | -43% | 0 | 0 | — |
case-09 | fail→pass | 6,919 | 2,404 | -65% | 1 | 1 | 0% | 1,041 | 801 | -23% | 0 | 0 | — |
case-10 | fail→pass | 7,258 | 4,053 | -44% | 1 | 1 | 0% | 1,134 | 852 | -25% | 0 | 0 | — |
case-11 | fail→pass | 15,583 | 7,974 | -49% | 1 | 1 | 0% | 2,454 | 1,831 | -25% | 0 | 0 | — |
case-12 | fail→pass | 9,576 | 2,279 | -76% | 1 | 1 | 0% | 1,755 | 816 | -54% | 0 | 0 | — |
case-13 | fail→fail | 14,713 | 3,608 | -75% | 1 | 1 | 0% | 2,554 | 1,053 | -59% | 0 | 0 | — |
case-14 | fail→pass | 11,264 | 2,409 | -79% | 1 | 1 | 0% | 1,747 | 889 | -49% | 0 | 0 | — |
case-15 | fail→fail | 30,832 | 6,263 | -80% | 1 | 1 | 0% | 824 | 1,589 | +93% | 0 | 0 | — |
case-16 | fail→pass | 6,194 | 2,879 | -54% | 1 | 1 | 0% | 994 | 902 | -9% | 0 | 0 | — |
case-17 | pass→pass | 8,112 | 4,152 | -49% | 1 | 1 | 0% | 1,249 | 1,166 | -7% | 0 | 0 | — |
case-18 | fail→pass | 11,722 | 2,104 | -82% | 1 | 1 | 0% | 1,850 | 744 | -60% | 0 | 0 | — |
case-19 | pass→pass | 8,118 | 3,679 | -55% | 1 | 1 | 0% | 1,199 | 961 | -20% | 0 | 0 | — |
case-20 | fail→pass | 17,221 | 6,129 | -64% | 1 | 1 | 0% | 2,550 | 1,501 | -41% | 0 | 0 | — |
case-21 | fail→pass | 8,232 | 4,747 | -42% | 1 | 1 | 0% | 1,199 | 1,182 | -1% | 0 | 0 | — |
case-22 | fail→pass | 13,493 | 3,022 | -78% | 1 | 1 | 0% | 2,013 | 939 | -53% | 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, and 20 counted toward the lift figure. The other 2 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 +64 percentage points is the difference between those two pass rates over the 20 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.