Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Japanese ISMAP (Information System Security Management and Assessment Program) expert. Provides guidance on ISO 27001/27017/27018 compliance, Japanese government cloud requirements, and data residency in Tokyo/Osaka regions.
.claude/skills/grcengclub-ismap-expert/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -28% | 0% |
Expertise in Japanese government cloud security program based on ISO 27001/27017/27018.
Authority: Digital Agency of Japan Base Standards:
Scope: Cloud services for Japanese government agencies
14 control domains for information security management.
Additional cloud-specific controls (CLD prefix) for providers and customers.
PII protection requirements for public cloud services.
Regions: ap-northeast-1 (Tokyo), ap-northeast-3 (Osaka) Requirement: Government data in Japanese regions only
Timeline: 6-12 months
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 17,351 | 22,120 | +27% | 1 | 1 | 0% | 2,725 | 2,458 | -10% | 0 | 0 | — |
case-02 | fail→pass | 18,459 | 16,089 | -13% | 1 | 1 | 0% | 2,821 | 2,984 | +6% | 0 | 0 | — |
case-03 | fail→pass | 11,663 | 9,205 | -21% | 1 | 1 | 0% | 2,006 | 2,011 | +0% | 0 | 0 | — |
case-04 | fail→pass | 18,645 | 7,073 | -62% | 1 | 1 | 0% | 1,272 | 1,695 | +33% | 0 | 0 | — |
case-05 | fail→pass | 11,480 | 5,468 | -52% | 1 | 1 | 0% | 1,747 | 1,217 | -30% | 0 | 0 | — |
case-06 | fail→pass | 8,425 | 4,379 | -48% | 1 | 1 | 0% | 1,461 | 1,059 | -28% | 0 | 0 | — |
case-07 | fail→pass | 15,705 | 5,157 | -67% | 1 | 1 | 0% | 950 | 1,220 | +28% | 0 | 0 | — |
case-08 | pass→pass | 6,722 | 3,375 | -50% | 1 | 1 | 0% | 1,043 | 881 | -16% | 0 | 0 | — |
case-09 | pass→pass | 6,429 | 4,886 | -24% | 1 | 1 | 0% | 1,041 | 1,083 | +4% | 0 | 0 | — |
case-10 | pass→pass | 11,302 | 12,103 | +7% | 1 | 1 | 0% | 1,713 | 2,282 | +33% | 0 | 0 | — |
case-11 | fail→pass | 14,035 | 11,373 | -19% | 1 | 1 | 0% | 2,097 | 2,092 | -0% | 0 | 0 | — |
case-12 | fail→pass | 17,480 | 15,167 | -13% | 1 | 1 | 0% | 2,667 | 2,719 | +2% | 0 | 0 | — |
case-13 | pass→pass | 19,281 | 16,428 | -15% | 1 | 1 | 0% | 3,136 | 3,014 | -4% | 0 | 0 | — |
case-14 | pass→pass | 18,641 | 15,061 | -19% | 1 | 1 | 0% | 2,985 | 2,965 | -1% | 0 | 0 | — |
case-15 | pass→pass | 14,719 | 9,797 | -33% | 1 | 1 | 0% | 2,455 | 2,057 | -16% | 0 | 0 | — |
case-16 | pass→pass | 4,863 | 3,258 | -33% | 1 | 1 | 0% | 814 | 900 | +11% | 0 | 0 | — |
case-17 | pass→pass | 6,487 | 2,008 | -69% | 1 | 1 | 0% | 1,200 | 678 | -44% | 0 | 0 | — |
case-18 | fail→pass | 17,060 | 14,154 | -17% | 1 | 1 | 0% | 2,424 | 2,491 | +3% | 0 | 0 | — |
case-19 | fail→pass | 5,941 | 2,691 | -55% | 1 | 1 | 0% | 893 | 760 | -15% | 0 | 0 | — |
case-20 | fail→pass | 5,919 | 3,098 | -48% | 1 | 1 | 0% | 915 | 826 | -10% | 0 | 0 | — |
case-21 | pass→pass | 6,656 | 5,233 | -21% | 1 | 1 | 0% | 972 | 1,144 | +18% | 0 | 0 | — |
case-22 | fail→pass | 6,030 | 3,503 | -42% | 1 | 1 | 0% | 974 | 952 | -2% | 0 | 0 | — |
case-23 | pass→pass | 3,461 | 2,400 | -31% | 1 | 1 | 0% | 491 | 700 | +43% | 0 | 0 | — |
case-24 | pass→pass | 7,585 | 2,988 | -61% | 1 | 1 | 0% | 1,073 | 796 | -26% | 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. 24 cases were attempted, and 22 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 +50 percentage points is the difference between those two pass rates over the 22 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.