Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Background legal and regulatory context for the shinkoku tax filing plugin. Contains the standard disclaimer about the scope of tax information provided, the relationship to the Tax Accountant Act (税理士法), and tool limitations. This skill is not user-invocable — Claude loads it automatically when generating tax-related responses that require a disclaimer.
.claude/skills/kazukinagata-tax-legal-context/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -21% | 0% |
このスキルは shinkoku の税務関連回答における法的・免責コンテキストを提供する。
免責事項が必要な回答を行う際は、references/disclaimer.md を読み込んで以下を実行する:
references/disclaimer.md の「標準免責文」を回答末尾に付記する| ファイル | 内容 | |---------|------| | references/disclaimer.md | 標準免責文・税理士法との関係・ツール制限事項・情報の正確性 |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 21,063 | 11,167 | -47% | 1 | 1 | 0% | 2,734 | 2,011 | -26% | 0 | 0 | — |
case-02 | pass→pass | 16,082 | 11,622 | -28% | 1 | 1 | 0% | 2,520 | 1,865 | -26% | 0 | 0 | — |
case-03 | pass→pass | 12,800 | 12,747 | -0% | 1 | 1 | 0% | 2,112 | 1,636 | -23% | 0 | 0 | — |
case-04 | pass→pass | 15,229 | 12,641 | -17% | 1 | 1 | 0% | 2,481 | 2,254 | -9% | 0 | 0 | — |
case-05 | fail→pass | 15,980 | 13,368 | -16% | 1 | 1 | 0% | 2,522 | 2,393 | -5% | 0 | 0 | — |
case-06 | pass→pass | 14,003 | 11,495 | -18% | 1 | 1 | 0% | 2,436 | 1,818 | -25% | 0 | 0 | — |
case-07 | fail→pass | 21,093 | 11,129 | -47% | 1 | 1 | 0% | 3,215 | 1,929 | -40% | 0 | 0 | — |
case-08 | fail→pass | 14,306 | 9,259 | -35% | 1 | 1 | 0% | 2,735 | 1,871 | -32% | 0 | 0 | — |
case-09 | pass→pass | 17,175 | 11,589 | -33% | 1 | 1 | 0% | 2,572 | 2,060 | -20% | 0 | 0 | — |
case-10 | pass→pass | 14,589 | 11,368 | -22% | 1 | 1 | 0% | 2,504 | 2,096 | -16% | 0 | 0 | — |
case-11 | fail→pass | 12,348 | 8,044 | -35% | 1 | 1 | 0% | 2,407 | 1,599 | -34% | 0 | 0 | — |
case-12 | pass→pass | 13,678 | 9,201 | -33% | 1 | 1 | 0% | 2,502 | 1,909 | -24% | 0 | 0 | — |
case-13 | pass→pass | 11,334 | 9,170 | -19% | 1 | 1 | 0% | 2,240 | 1,696 | -24% | 0 | 0 | — |
case-14 | pass→pass | 12,695 | 6,114 | -52% | 1 | 1 | 0% | 2,206 | 1,156 | -48% | 0 | 0 | — |
case-15 | fail→pass | 9,166 | 6,329 | -31% | 1 | 1 | 0% | 1,798 | 1,426 | -21% | 0 | 0 | — |
case-16 | pass→pass | 13,259 | 8,676 | -35% | 1 | 1 | 0% | 2,596 | 1,493 | -42% | 0 | 0 | — |
case-17 | pass→pass | 14,237 | 10,939 | -23% | 1 | 1 | 0% | 2,538 | 2,195 | -14% | 0 | 0 | — |
case-18 | pass→pass | 14,948 | 11,599 | -22% | 1 | 1 | 0% | 2,758 | 2,168 | -21% | 0 | 0 | — |
case-19 | pass→pass | 13,711 | 12,506 | -9% | 1 | 1 | 0% | 2,535 | 2,214 | -13% | 0 | 0 | — |
case-20 | pass→fail | 15,235 | 4,057 | -73% | 1 | 1 | 0% | 2,421 | 591 | -76% | 0 | 0 | — |
case-21 | pass→pass | 3,723 | 3,962 | +6% | 1 | 1 | 0% | 749 | 906 | +21% | 0 | 0 | — |
case-22 | fail→pass | 8,862 | 2,958 | -67% | 1 | 1 | 0% | 1,746 | 708 | -59% | 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. The headline lift of +23 percentage points is the difference between those two pass rates over the 22 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.