Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Specialized skill for Zephyr RTOS development and configuration
.claude/skills/a5c-ai-zephyr-rtos/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-15 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-13 | ✓→✗ | ▼ Worse | 14% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | -5% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 59% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 33% | 0% |
This skill provides specialized support for Zephyr RTOS development, including device tree configuration, Kconfig management, and integration with Zephyr's extensive subsystems.
rtos-integration.js - Zephyr RTOS setup and integrationsecure-boot-implementation.js - MCUboot and secure bootota-firmware-update.js - Firmware update with MCUbootlow-power-design.js - Zephyr power managementThis skill is invoked when tasks require:
app/
CMakeLists.txt
prj.conf
boards/
my_board.overlay
src/
main.cdts&i2c0 { status = "okay"; clock-frequency = <I2C_BITRATE_FAST>; sensor@48 { compatible = "ti,tmp102"; reg = <0x48>; }; };
kconfigCONFIG_BT=y CONFIG_BT_PERIPHERAL=y CONFIG_BT_DEVICE_NAME="My Device" CONFIG_BT_GATT_DYNAMIC_DB=y
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 10,269 | 37,196 | +262% | 1 | 1 | 0% | 1,932 | 1,827 | -5% | 0 | 0 | — |
case-02 | pass→pass | 9,377 | 10,463 | +12% | 1 | 1 | 0% | 1,370 | 2,173 | +59% | 0 | 0 | — |
case-03 | pass→pass | 16,179 | 15,053 | -7% | 1 | 1 | 0% | 2,430 | 3,231 | +33% | 0 | 0 | — |
case-04 | pass→pass | 7,267 | 4,815 | -34% | 1 | 1 | 0% | 1,084 | 1,477 | +36% | 0 | 0 | — |
case-05 | pass→pass | 17,833 | 15,629 | -12% | 1 | 1 | 0% | 2,839 | 3,030 | +7% | 0 | 0 | — |
case-06 | pass→pass | 6,904 | 9,797 | +42% | 1 | 1 | 0% | 1,592 | 2,108 | +32% | 0 | 0 | — |
case-07 | pass→pass | 9,985 | 10,307 | +3% | 1 | 1 | 0% | 1,785 | 2,560 | +43% | 0 | 0 | — |
case-08 | pass→pass | 8,811 | 10,589 | +20% | 1 | 1 | 0% | 1,679 | 2,187 | +30% | 0 | 0 | — |
case-09 | pass→pass | 13,198 | 10,786 | -18% | 1 | 1 | 0% | 1,763 | 2,724 | +55% | 0 | 0 | — |
case-10 | pass→pass | 11,013 | 9,369 | -15% | 1 | 1 | 0% | 2,192 | 2,532 | +16% | 0 | 0 | — |
case-11 | pass→pass | 13,700 | 12,345 | -10% | 1 | 1 | 0% | 2,606 | 2,982 | +14% | 0 | 0 | — |
case-12 | pass→pass | 20,701 | 11,551 | -44% | 1 | 1 | 0% | 3,378 | 3,053 | -10% | 0 | 0 | — |
case-13 | pass→fail | 10,674 | 10,146 | -5% | 1 | 1 | 0% | 1,895 | 2,161 | +14% | 0 | 0 | — |
case-14 | pass→pass | 10,920 | 14,212 | +30% | 1 | 1 | 0% | 2,003 | 2,799 | +40% | 0 | 0 | — |
case-15 | fail→pass | 15,033 | 12,207 | -19% | 1 | 1 | 0% | 2,413 | 3,012 | +25% | 0 | 0 | — |
case-16 | pass→pass | 16,105 | 19,379 | +20% | 1 | 1 | 0% | 2,515 | 3,578 | +42% | 0 | 0 | — |
case-17 | pass→pass | 6,827 | 6,409 | -6% | 1 | 1 | 0% | 1,281 | 1,695 | +32% | 0 | 0 | — |
case-18 | pass→pass | 10,700 | 12,429 | +16% | 1 | 1 | 0% | 2,109 | 2,672 | +27% | 0 | 0 | — |
case-19 | pass→pass | 18,149 | 11,534 | -36% | 1 | 1 | 0% | 1,496 | 2,454 | +64% | 0 | 0 | — |
case-20 | pass→pass | 7,033 | 9,697 | +38% | 1 | 1 | 0% | 1,451 | 2,130 | +47% | 0 | 0 | — |
case-21 | pass→pass | 10,936 | 9,811 | -10% | 1 | 1 | 0% | 1,947 | 2,101 | +8% | 0 | 0 | — |
case-22 | pass→pass | 21,753 | 16,514 | -24% | 1 | 1 | 0% | 3,389 | 3,186 | -6% | 0 | 0 | — |
case-23 | pass→pass | 12,955 | 10,539 | -19% | 1 | 1 | 0% | 1,885 | 2,170 | +15% | 0 | 0 | — |
case-24 | pass→pass | 20,141 | 20,466 | +2% | 1 | 1 | 0% | 3,381 | 4,028 | +19% | 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. The headline lift of 0 percentage points is the difference between those two pass rates over the 24 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.