Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Codex's capabilities with specialized knowledge, workflows, or tool integrations.
.claude/skills/bilal140202-skill-creator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -18% | 0% |
Create a new skill by generating a SKILL.md file with YAML frontmatter.
my-skill/
├── SKILL.md # Instructions + frontmatter
├── template.md # (optional) Template files
└── example.py # (optional) Example scriptsmarkdown--- name: skill-name description: "Brief description of what this skill does" tags: [category1, category2] version: "1.0.0" --- # Skill Name ## Purpose What this skill accomplishes. ## Instructions Step-by-step instructions for the agent. ## Examples Concrete examples of input/output.
api-client-generator)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | fail→fail | 19,030 | 13,611 | -28% | 1 | 1 | 0% | 3,729 | 2,979 | -20% | 0 | 0 | — |
case-01 | fail→pass | 14,181 | 9,642 | -32% | 1 | 1 | 0% | 2,749 | 2,258 | -18% | 0 | 0 | — |
case-02 | fail→pass | 16,435 | 11,273 | -31% | 1 | 1 | 0% | 2,940 | 2,618 | -11% | 0 | 0 | — |
case-14 | fail→pass | 12,003 | 8,449 | -30% | 1 | 1 | 0% | 2,377 | 2,184 | -8% | 0 | 0 | — |
case-03 | fail→pass | 29,657 | 12,597 | -58% | 1 | 1 | 0% | 2,671 | 2,500 | -6% | 0 | 0 | — |
case-04 | fail→fail | 26,630 | 11,164 | -58% | 1 | 1 | 0% | 1,374 | 2,832 | +106% | 0 | 0 | — |
case-05 | pass→pass | 9,847 | 9,730 | -1% | 1 | 1 | 0% | 2,227 | 2,562 | +15% | 0 | 0 | — |
case-06 | pass→pass | 11,983 | 8,512 | -29% | 1 | 1 | 0% | 2,327 | 1,826 | -22% | 0 | 0 | — |
case-07 | fail→fail | 11,342 | 13,018 | +15% | 1 | 1 | 0% | 2,380 | 2,895 | +22% | 0 | 0 | — |
case-08 | fail→fail | 11,454 | 10,369 | -9% | 1 | 1 | 0% | 2,220 | 2,453 | +10% | 0 | 0 | — |
case-09 | fail→fail | 8,181 | 6,881 | -16% | 1 | 1 | 0% | 1,707 | 1,822 | +7% | 0 | 0 | — |
case-10 | fail→fail | 9,983 | 9,428 | -6% | 1 | 1 | 0% | 1,929 | 2,285 | +18% | 0 | 0 | — |
case-11 | fail→fail | 8,481 | 7,634 | -10% | 1 | 1 | 0% | 1,865 | 1,785 | -4% | 0 | 0 | — |
case-12 | fail→pass | 11,017 | 7,838 | -29% | 1 | 1 | 0% | 2,331 | 1,902 | -18% | 0 | 0 | — |
case-15 | pass→pass | 12,179 | 11,898 | -2% | 1 | 1 | 0% | 2,145 | 2,655 | +24% | 0 | 0 | — |
case-16 | fail→pass | 18,731 | 11,566 | -38% | 1 | 1 | 0% | 3,764 | 2,759 | -27% | 0 | 0 | — |
case-17 | fail→pass | 11,678 | 10,485 | -10% | 1 | 1 | 0% | 2,600 | 2,266 | -13% | 0 | 0 | — |
case-18 | fail→fail | 17,715 | 18,473 | +4% | 1 | 1 | 0% | 3,325 | 3,995 | +20% | 0 | 0 | — |
case-19 | fail→pass | 15,568 | 10,592 | -32% | 1 | 1 | 0% | 3,582 | 2,351 | -34% | 0 | 0 | — |
case-20 | pass→pass | 9,165 | 7,980 | -13% | 1 | 1 | 0% | 2,101 | 1,875 | -11% | 0 | 0 | — |
case-21 | fail→pass | 13,410 | 13,076 | -2% | 1 | 1 | 0% | 2,705 | 3,458 | +28% | 0 | 0 | — |
case-22 | fail→pass | 15,888 | 13,149 | -17% | 1 | 1 | 0% | 2,992 | 2,767 | -8% | 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 21 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 +45 percentage points is the difference between those two pass rates over the 21 comparable cases.
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.