Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate and edit images using OpenAI's Image API for project assets — UI mockups, icons, illustrations, social cards, and visual references.
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -75% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 31% | 0% |
| case-15 | ✓→✗ | ▼ Worse | -87% | 0% |
| case-20 | ✓→✗ | ▼ Worse | -78% | 0% |
> Curated from OpenAI's skills repository.
Generate and edit images using OpenAI's Image API for project assets — UI mockups, icons, illustrations, social cards, and visual references.
image-generationThis catalogue entry advertises the skill in OpenDesign so the agent discovers it during planning. To run the full upstream workflow with its original assets, scripts, and references, install the upstream bundle into your active agent's skills directory:
bash# Inspect the upstream README for exact paths open https://github.com/openai/skills
Then ask the agent to invoke this skill by name (imagegen) or with one of the trigger phrases listed in this skill's frontmatter.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,071 | 38,589 | +102% | 1 | 1 | 0% | 3,772 | 1,508 | -60% | 0 | 0 | — |
case-02 | fail→fail | 16,528 | 10,812 | -35% | 1 | 1 | 0% | 3,195 | 672 | -79% | 0 | 0 | — |
case-03 | fail→pass | 15,080 | 8,329 | -45% | 1 | 1 | 0% | 3,026 | 1,651 | -45% | 0 | 0 | — |
case-04 | pass→pass | 7,588 | 5,283 | -30% | 1 | 1 | 0% | 1,340 | 1,125 | -16% | 0 | 0 | — |
case-05 | fail→pass | 17,026 | 4,325 | -75% | 1 | 1 | 0% | 3,373 | 846 | -75% | 0 | 0 | — |
case-06 | fail→fail | 15,420 | 8,493 | -45% | 1 | 1 | 0% | 2,978 | 708 | -76% | 0 | 0 | — |
case-07 | fail→fail | 16,656 | 5,516 | -67% | 1 | 1 | 0% | 3,164 | 744 | -76% | 0 | 0 | — |
case-08 | fail→fail | 35,143 | 2,851 | -92% | 1 | 1 | 0% | 6,169 | 540 | -91% | 0 | 0 | — |
case-09 | fail→fail | 12,033 | 9,776 | -19% | 1 | 1 | 0% | 1,713 | 835 | -51% | 0 | 0 | — |
case-10 | fail→fail | 19,038 | 9,195 | -52% | 1 | 1 | 0% | 3,479 | 714 | -79% | 0 | 0 | — |
case-11 | fail→fail | 3,917 | 7,142 | +82% | 1 | 1 | 0% | 513 | 501 | -2% | 0 | 0 | — |
case-12 | fail→fail | 13,655 | 4,810 | -65% | 1 | 1 | 0% | 1,983 | 838 | -58% | 0 | 0 | — |
case-13 | fail→fail | 27,334 | 9,501 | -65% | 1 | 1 | 0% | 5,649 | 727 | -87% | 0 | 0 | — |
case-14 | fail→fail | 7,606 | 9,343 | +23% | 1 | 1 | 0% | 1,179 | 1,554 | +32% | 0 | 0 | — |
case-15 | pass→fail | 30,616 | 4,129 | -87% | 1 | 1 | 0% | 6,163 | 829 | -87% | 0 | 0 | — |
case-16 | fail→fail | 17,129 | 7,900 | -54% | 1 | 1 | 0% | 3,452 | 636 | -82% | 0 | 0 | — |
case-17 | fail→fail | 10,741 | 8,264 | -23% | 1 | 1 | 0% | 1,597 | 567 | -64% | 0 | 0 | — |
case-18 | fail→fail | 13,449 | 4,010 | -70% | 1 | 1 | 0% | 1,953 | 696 | -64% | 0 | 0 | — |
case-19 | fail→fail | 8,715 | 8,465 | -3% | 1 | 1 | 0% | 1,320 | 1,442 | +9% | 0 | 0 | — |
case-20 | pass→fail | 19,388 | 10,717 | -45% | 1 | 1 | 0% | 4,100 | 916 | -78% | 0 | 0 | — |
case-21 | pass→fail | 6,078 | 8,559 | +41% | 1 | 1 | 0% | 902 | 650 | -28% | 0 | 0 | — |
case-22 | fail→pass | 3,414 | 10,021 | +194% | 1 | 1 | 0% | 502 | 659 | +31% | 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 14 counted toward the lift figure. The other 8 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 0 percentage points is the difference between those two pass rates over the 14 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.