Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A test skill for parity testing.
.claude/skills/bilal140202-myskill/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -54% | 0% |
Test skill for parity testing. Validates that the Go port correctly identifies well-formed SKILL.md files with all required sections.
This skill applies when running validate parity tests.
Run against the fixture directory with yakos validate <path>. Expected result: [ok] for this file, Summary: 0 error(s).
yakos validate for this fixture.[err] or [warn] lines appear for this skill file.--all flag and verify the same result.find order in bash vs WalkDir order in Go differ; paritytests sort both outputs before comparing.
80-350, rules: 60-150) to avoid warnings.
INDEX.md and README.md files are excluded from frontmattervalidation in all three subdirectories.
because they all contain "/lib/" in their path.
does full YAML parsing via gopkg.in/yaml.v3).
equivalent in Go.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 19,303 | 8,979 | -53% | 1 | 1 | 0% | 3,309 | 2,015 | -39% | 0 | 0 | — |
case-02 | fail→pass | 17,700 | 10,795 | -39% | 1 | 1 | 0% | 2,896 | 2,409 | -17% | 0 | 0 | — |
case-03 | fail→pass | 17,077 | 8,860 | -48% | 1 | 1 | 0% | 2,951 | 2,006 | -32% | 0 | 0 | — |
case-04 | fail→pass | 21,842 | 1,849 | -92% | 1 | 1 | 0% | 1,644 | 696 | -58% | 0 | 0 | — |
case-05 | fail→pass | 10,947 | 2,296 | -79% | 1 | 1 | 0% | 1,576 | 721 | -54% | 0 | 0 | — |
case-06 | fail→pass | 20,790 | 1,943 | -91% | 1 | 1 | 0% | 918 | 715 | -22% | 0 | 0 | — |
case-07 | fail→pass | 10,841 | 2,046 | -81% | 1 | 1 | 0% | 1,716 | 714 | -58% | 0 | 0 | — |
case-08 | fail→pass | 16,451 | 3,748 | -77% | 1 | 1 | 0% | 2,552 | 1,017 | -60% | 0 | 0 | — |
case-09 | fail→pass | 11,425 | 1,637 | -86% | 1 | 1 | 0% | 1,861 | 672 | -64% | 0 | 0 | — |
case-10 | pass→pass | 6,521 | 2,622 | -60% | 1 | 1 | 0% | 1,020 | 861 | -16% | 0 | 0 | — |
case-11 | fail→pass | 7,693 | 1,541 | -80% | 1 | 1 | 0% | 1,139 | 613 | -46% | 0 | 0 | — |
case-12 | fail→pass | 20,112 | 1,529 | -92% | 1 | 1 | 0% | 862 | 620 | -28% | 0 | 0 | — |
case-13 | fail→pass | 15,632 | 2,499 | -84% | 1 | 1 | 0% | 2,488 | 779 | -69% | 0 | 0 | — |
case-14 | pass→pass | 3,232 | 1,605 | -50% | 1 | 1 | 0% | 502 | 662 | +32% | 0 | 0 | — |
case-15 | fail→pass | 48,285 | 2,208 | -95% | 1 | 1 | 0% | 356 | 819 | +130% | 0 | 0 | — |
case-16 | pass→pass | 7,565 | 6,314 | -17% | 1 | 1 | 0% | 1,251 | 1,419 | +13% | 0 | 0 | — |
case-17 | pass→pass | 12,537 | 8,255 | -34% | 1 | 1 | 0% | 2,081 | 1,981 | -5% | 0 | 0 | — |
case-18 | pass→pass | 7,284 | 5,706 | -22% | 1 | 1 | 0% | 1,472 | 1,411 | -4% | 0 | 0 | — |
case-19 | pass→pass | 11,670 | 10,522 | -10% | 1 | 1 | 0% | 2,188 | 2,446 | +12% | 0 | 0 | — |
case-20 | fail→pass | 7,255 | 1,753 | -76% | 1 | 1 | 0% | 1,224 | 650 | -47% | 0 | 0 | — |
case-21 | fail→pass | 18,445 | 1,553 | -92% | 1 | 1 | 0% | 1,022 | 644 | -37% | 0 | 0 | — |
case-22 | fail→pass | 3,382 | 2,295 | -32% | 1 | 1 | 0% | 487 | 802 | +65% | 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 18 counted toward the lift figure. The other 4 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 +73 percentage points is the difference between those two pass rates over the 18 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.