Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →IEC 62304 medical device software lifecycle framework. Risk classification, software safety classification (A/B/C), design control, verification, validation, and regulatory documentation for FDA/EU MDR submissions.
.claude/skills/mkurman-iec-62304/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-16 | ✓→✓ | ＝ Same ✓ | 36% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | -10% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 46% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 27% | 0% |
IEC 62304 medical device software lifecycle framework. Software safety classification (Class A/B/C), design control, verification, validation, and regulatory documentation for FDA and EU MDR submissions.
- Software Development Plan
- Software Requirements Specification
- Software Architecture Document
- Software Detailed Design
- Software Unit Verification
- Software Integration Testing
- Software System Testing
- Software Release Notes
- Software Maintenance Plan
- Software Risk Management File| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-16 | pass→pass | 14,098 | 15,897 | +13% | 1 | 1 | 0% | 2,313 | 3,149 | +36% | 0 | 0 | — |
case-01 | pass→pass | 14,907 | 11,619 | -22% | 1 | 1 | 0% | 2,348 | 2,112 | -10% | 0 | 0 | — |
case-02 | fail→pass | 15,220 | 7,986 | -48% | 1 | 1 | 0% | 2,480 | 1,603 | -35% | 0 | 0 | — |
case-03 | pass→pass | 8,923 | 11,291 | +27% | 1 | 1 | 0% | 1,468 | 2,139 | +46% | 0 | 0 | — |
case-04 | pass→pass | 5,892 | 6,732 | +14% | 1 | 1 | 0% | 1,070 | 1,357 | +27% | 0 | 0 | — |
case-05 | pass→pass | 6,068 | 3,868 | -36% | 1 | 1 | 0% | 990 | 858 | -13% | 0 | 0 | — |
case-06 | pass→pass | 15,344 | 17,108 | +11% | 1 | 1 | 0% | 2,614 | 3,225 | +23% | 0 | 0 | — |
case-07 | pass→pass | 6,530 | 4,423 | -32% | 1 | 1 | 0% | 1,092 | 978 | -10% | 0 | 0 | — |
case-08 | pass→pass | 16,770 | 16,447 | -2% | 1 | 1 | 0% | 2,852 | 3,013 | +6% | 0 | 0 | — |
case-09 | pass→pass | 14,994 | 15,657 | +4% | 1 | 1 | 0% | 2,359 | 2,837 | +20% | 0 | 0 | — |
case-10 | pass→pass | 11,824 | 7,577 | -36% | 1 | 1 | 0% | 2,002 | 1,703 | -15% | 0 | 0 | — |
case-11 | pass→pass | 13,038 | 15,340 | +18% | 1 | 1 | 0% | 2,414 | 2,751 | +14% | 0 | 0 | — |
case-12 | pass→pass | 15,143 | 11,944 | -21% | 1 | 1 | 0% | 2,655 | 2,484 | -6% | 0 | 0 | — |
case-13 | pass→pass | 7,864 | 7,285 | -7% | 1 | 1 | 0% | 1,343 | 1,509 | +12% | 0 | 0 | — |
case-14 | pass→pass | 7,251 | 7,373 | +2% | 1 | 1 | 0% | 1,187 | 1,393 | +17% | 0 | 0 | — |
case-15 | pass→pass | 5,390 | 3,563 | -34% | 1 | 1 | 0% | 1,003 | 906 | -10% | 0 | 0 | — |
case-17 | pass→pass | 13,007 | 10,726 | -18% | 1 | 1 | 0% | 2,293 | 2,147 | -6% | 0 | 0 | — |
case-18 | pass→pass | 13,272 | 10,422 | -21% | 1 | 1 | 0% | 2,313 | 2,047 | -12% | 0 | 0 | — |
case-19 | pass→pass | 12,788 | 8,401 | -34% | 1 | 1 | 0% | 2,084 | 1,858 | -11% | 0 | 0 | — |
case-20 | pass→pass | 16,802 | 15,808 | -6% | 1 | 1 | 0% | 2,750 | 2,991 | +9% | 0 | 0 | — |
case-21 | pass→pass | 17,834 | 22,590 | +27% | 1 | 1 | 0% | 3,010 | 3,998 | +33% | 0 | 0 | — |
case-22 | pass→pass | 18,939 | 16,880 | -11% | 1 | 1 | 0% | 3,058 | 3,046 | -0% | 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 +5 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.