Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Analyze Taiwan labor law fundamentals under the Labor Standards Act including working hours, overtime, leave, and termination rules. Use this skill when the user needs to understand employment obligations, calculate overtime pay, evaluate whether a termination is lawful, or design compliant HR policies — even if they say 'how much overtime pay do I owe', 'can I fire this employee', 'what leave are employees entitled to', or 'are we complying with labor law'.
.claude/skills/asgard-ai-platform-law-labor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 62% | 0% |
The Labor Standards Act (勞基法, LSA) sets minimum labor standards in Taiwan. It covers working hours, overtime, leave, wages, and termination. These are MINIMUMS — employers can offer better but not worse terms. This skill focuses on Taiwan-specific requirements.
IRON LAW: LSA Is a Floor, Not a Ceiling
Everything in the LSA represents the MINIMUM employer obligation.
Employment contracts can exceed these standards but NEVER go below them.
A contract clause that provides less than LSA minimums is void even if
the employee signed it. "The employee agreed to it" is not a defense.| Rule | Standard | |------|---------| | Regular hours | 8 hours/day, 40 hours/week | | Overtime cap | 46 hours/month (can extend to 54 with labor-management agreement) | | Rest day | 1 regular day off (例假) + 1 rest day (休息日) per week |
Overtime Pay Calculation: | Period | Rate | |--------|------| | Weekday overtime, first 2 hours | 1.34x regular hourly rate | | Weekday overtime, hours 3-4 | 1.67x regular hourly rate | | Rest day (休息日) overtime, first 2 hours | 1.34x (minimum 4 hours counted) | | Rest day overtime, hours 3-8 | 1.67x | | Rest day overtime, hours 9-12 | 2.67x | | Regular day off (例假) overtime | 2x + compensatory day off | | National holiday overtime | 2x |
| Type | Entitlement | |------|------------| | Annual leave (特休) | 6 months: 3 days → 1 year: 7 days → 2 years: 10 days → 3-5 years: 14 days → 5-10 years: 15 days → 10+ years: +1 day per year (max 30) | | Sick leave (病假) | 30 days/year at half pay (hospitalization: additional 1 year) | | Personal leave (事假) | 14 days/year, unpaid | | Maternity leave (產假) | 8 weeks at full pay (after 6 months employment) | | Paternity leave (陪產假) | 7 days at full pay | | Menstrual leave (生理假) | 1 day/month (first 3 counted as sick leave at half pay; additional at no pay) |
Employer can terminate WITH notice (Art. 11):
Notice periods: | Tenure | Notice Required | |--------|----------------| | 3 months - 1 year | 10 days | | 1 - 3 years | 20 days | | 3+ years | 30 days |
Employer can terminate WITHOUT notice (Art. 12):
Severance pay: 0.5 months' average wage per year of service (new system under Labor Pension Act)
markdown# Labor Law Analysis: {Situation} ## Applicable Rules | Rule | LSA Article | Requirement | |------|-----------|-------------| | {topic} | Art. {N} | {requirement} | ## Compliance Assessment | Area | Status | Gap | |------|--------|-----| | Working hours | ✓/✗ | {detail} | | Overtime pay | ✓/✗ | {detail} | | Leave | ✓/✗ | {detail} | | Termination | ✓/✗ | {detail} | ## Recommendation {Specific actions to achieve compliance}
Scenario: Employee works 10 hours on a Tuesday (regular workday). Base monthly salary: NT$50,000.
Calculation:
references/lsa-full-text.mdreferences/labor-pension.md| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 25,577 | 24,000 | -6% | 1 | 1 | 0% | 4,509 | 4,546 | +1% | 0 | 0 | — |
case-02 | fail→pass | 16,864 | 16,797 | -0% | 1 | 1 | 0% | 3,167 | 4,380 | +38% | 0 | 0 | — |
case-03 | fail→pass | 22,511 | 15,457 | -31% | 1 | 1 | 0% | 4,299 | 4,333 | +1% | 0 | 0 | — |
case-04 | fail→fail | 19,665 | 22,677 | +15% | 1 | 1 | 0% | 4,122 | 6,013 | +46% | 0 | 0 | — |
case-05 | pass→pass | 17,713 | 21,020 | +19% | 1 | 1 | 0% | 2,780 | 4,537 | +63% | 0 | 0 | — |
case-06 | pass→pass | 17,171 | 16,872 | -2% | 1 | 1 | 0% | 3,017 | 4,625 | +53% | 0 | 0 | — |
case-07 | pass→pass | 22,937 | 28,622 | +25% | 1 | 1 | 0% | 4,694 | 5,479 | +17% | 0 | 0 | — |
case-08 | pass→pass | 29,160 | 18,386 | -37% | 1 | 1 | 0% | 6,093 | 4,694 | -23% | 0 | 0 | — |
case-09 | pass→pass | 16,973 | 13,245 | -22% | 1 | 1 | 0% | 2,881 | 3,478 | +21% | 0 | 0 | — |
case-10 | pass→pass | 18,862 | 26,184 | +39% | 1 | 1 | 0% | 2,946 | 5,620 | +91% | 0 | 0 | — |
case-11 | pass→pass | 13,793 | 9,714 | -30% | 1 | 1 | 0% | 2,119 | 3,193 | +51% | 0 | 0 | — |
case-12 | fail→pass | 18,740 | 10,787 | -42% | 1 | 1 | 0% | 2,978 | 3,159 | +6% | 0 | 0 | — |
case-13 | pass→pass | 17,418 | 10,572 | -39% | 1 | 1 | 0% | 2,109 | 3,314 | +57% | 0 | 0 | — |
case-14 | pass→pass | 10,386 | 7,983 | -23% | 1 | 1 | 0% | 1,971 | 2,850 | +45% | 0 | 0 | — |
case-15 | pass→pass | 15,471 | 7,054 | -54% | 1 | 1 | 0% | 2,400 | 2,721 | +13% | 0 | 0 | — |
case-16 | fail→fail | 16,200 | 13,017 | -20% | 1 | 1 | 0% | 2,205 | 3,837 | +74% | 0 | 0 | — |
case-17 | pass→pass | 13,341 | 9,764 | -27% | 1 | 1 | 0% | 2,177 | 3,305 | +52% | 0 | 0 | — |
case-18 | fail→pass | 13,842 | 13,726 | -1% | 1 | 1 | 0% | 2,169 | 3,517 | +62% | 0 | 0 | — |
case-19 | pass→pass | 12,868 | 11,855 | -8% | 1 | 1 | 0% | 2,461 | 3,584 | +46% | 0 | 0 | — |
case-20 | pass→pass | 14,811 | 9,480 | -36% | 1 | 1 | 0% | 2,582 | 3,046 | +18% | 0 | 0 | — |
case-21 | pass→pass | 13,951 | 9,784 | -30% | 1 | 1 | 0% | 2,254 | 3,131 | +39% | 0 | 0 | — |
case-22 | pass→pass | 15,804 | 12,838 | -19% | 1 | 1 | 0% | 2,590 | 3,547 | +37% | 0 | 0 | — |
case-23 | fail→pass | 14,217 | 14,617 | +3% | 1 | 1 | 0% | 2,243 | 3,609 | +61% | 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. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 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.