Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Deep expertise in ARM Cortex-M architecture and peripherals
.claude/skills/a5c-ai-arm-cortex-m/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-20 | ✓→✗ | ▼ Worse | 40% | 0% |
| case-01 | ✓→✓ | = Same ✓ | -26% | 0% |
| case-02 | ✓→✓ | = Same ✓ | 2% | 0% |
| case-03 | ✓→✓ | = Same ✓ | 13% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 87% | 0% |
This skill provides deep expertise in ARM Cortex-M architecture, including core configuration, peripheral programming, and low-level optimization for the most widely used embedded processor family.
bsp-development.js - BSP with Cortex-M supportisr-design.js - Interrupt architecture designmemory-architecture-planning.js - Memory layout with MPUreal-time-architecture-design.js - Real-time Cortex-M designbootloader-implementation.js - Cortex-M bootloaderThis skill is invoked when tasks require:
| Core | Features | |------|----------| | Cortex-M0/M0+ | Minimal, low-power | | Cortex-M3 | Full Thumb-2, MPU optional | | Cortex-M4 | DSP, optional FPU | | Cortex-M7 | Cache, dual-issue | | Cortex-M23 | TrustZone-M, security | | Cortex-M33 | TrustZone-M, DSP | | Cortex-M55 | MVE (Helium), ML |
cNVIC_SetPriorityGrouping(3); // 4 bits preemption, 0 bits sub NVIC_SetPriority(USART1_IRQn, NVIC_EncodePriority(3, 2, 0)); NVIC_EnableIRQ(USART1_IRQn);
cMPU->RNR = 0; // Region 0 MPU->RBAR = 0x20000000; // Base address MPU->RASR = MPU_RASR_ENABLE_Msk | (0x0F << MPU_RASR_SIZE_Pos) | // 64KB MPU_RASR_C_Msk | MPU_RASR_S_Msk;
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 16,961 | 13,700 | -19% | 1 | 1 | 0% | 3,547 | 2,642 | -26% | 0 | 0 | — |
case-02 | pass→pass | 21,753 | 19,985 | -8% | 1 | 1 | 0% | 3,806 | 3,890 | +2% | 0 | 0 | — |
case-03 | pass→pass | 20,570 | 13,280 | -35% | 1 | 1 | 0% | 3,082 | 3,494 | +13% | 0 | 0 | — |
case-04 | pass→pass | 6,311 | 7,668 | +22% | 1 | 1 | 0% | 1,272 | 2,381 | +87% | 0 | 0 | — |
case-05 | pass→pass | 20,806 | 19,946 | -4% | 1 | 1 | 0% | 3,084 | 3,726 | +21% | 0 | 0 | — |
case-06 | pass→pass | 18,428 | 11,934 | -35% | 1 | 1 | 0% | 3,019 | 3,447 | +14% | 0 | 0 | — |
case-07 | pass→pass | 10,314 | 9,511 | -8% | 1 | 1 | 0% | 991 | 1,547 | +56% | 0 | 0 | — |
case-08 | pass→pass | 9,332 | 5,840 | -37% | 1 | 1 | 0% | 742 | 1,763 | +138% | 0 | 0 | — |
case-09 | pass→pass | 19,591 | 19,808 | +1% | 1 | 1 | 0% | 2,766 | 3,659 | +32% | 0 | 0 | — |
case-10 | pass→pass | 25,406 | 28,372 | +12% | 1 | 1 | 0% | 3,782 | 4,649 | +23% | 0 | 0 | — |
case-11 | pass→pass | 11,999 | 6,544 | -45% | 1 | 1 | 0% | 1,469 | 1,955 | +33% | 0 | 0 | — |
case-12 | pass→pass | 7,903 | 8,491 | +7% | 1 | 1 | 0% | 495 | 1,391 | +181% | 0 | 0 | — |
case-13 | pass→pass | 8,666 | 9,586 | +11% | 1 | 1 | 0% | 670 | 1,652 | +147% | 0 | 0 | — |
case-14 | pass→pass | 19,992 | 22,336 | +12% | 1 | 1 | 0% | 3,650 | 4,796 | +31% | 0 | 0 | — |
case-15 | pass→pass | 8,986 | 10,514 | +17% | 1 | 1 | 0% | 874 | 1,811 | +107% | 0 | 0 | — |
case-16 | pass→pass | 7,723 | 11,918 | +54% | 1 | 1 | 0% | 1,335 | 1,827 | +37% | 0 | 0 | — |
case-17 | pass→pass | 27,450 | 19,617 | -29% | 1 | 1 | 0% | 4,495 | 3,647 | -19% | 0 | 0 | — |
case-18 | pass→pass | 15,593 | 16,356 | +5% | 1 | 1 | 0% | 2,007 | 3,004 | +50% | 0 | 0 | — |
case-19 | pass→pass | 21,299 | 29,079 | +37% | 1 | 1 | 0% | 3,185 | 5,401 | +70% | 0 | 0 | — |
case-20 | pass→fail | 12,368 | 11,394 | -8% | 1 | 1 | 0% | 1,356 | 1,892 | +40% | 0 | 0 | — |
case-21 | pass→pass | 26,651 | 25,944 | -3% | 1 | 1 | 0% | 4,516 | 5,306 | +17% | 0 | 0 | — |
case-22 | pass→pass | 15,214 | 11,294 | -26% | 1 | 1 | 0% | 2,094 | 2,947 | +41% | 0 | 0 | — |
case-23 | pass→pass | 15,205 | 15,791 | +4% | 1 | 1 | 0% | 1,956 | 2,796 | +43% | 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 -100 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Other measured skills in the registry, with their headline benchmark lift.