Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Iterate on the Cardputer-Adv MicroPython app bundle (Claude Buddy, Snake, Hello) after the device is already provisioned via m5-onboard. Use when the user wants to add a new app, push a single changed .py without re-flashing, watch device serial logs, or run a one-shot REPL command. Trigger on "add an app", "push to the cardputer", "tail the device", "run on the device", or follow-up work after /maker-setup.
.claude/skills/anthropics-cardputer-buddy/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -71% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -69% | 0% |
The buddy/ directory in the local build-with-claude clone is the MicroPython payload that m5-onboard installs onto /flash/. Work inside that clone.
/flash/
├── main.py launcher menu (replaces UIFlow's boot flow)
├── buddy_*.py shared libs (BLE, UI, state, protocol, chars)
├── burst_frames.py sprite frames
└── apps/
├── claude_buddy.py BLE client → Claude Desktop's Hardware Buddy
├── hello_cardputer.py
└── snake.pymain.py scans /flash/apps/ at boot and lists every .py as a menu entry. Drop a file into buddy/device/apps/, push it, and it appears on next boot.
Crib from buddy/device/apps/hello_cardputer.py — smallest example of keyboard polling, font, and exit conventions. Then push without re-flashing:
bashpython3 onboard/scripts/install_apps.py --port <PORT> --src buddy
<PORT> is whatever detect.py reported last run (e.g. /dev/cu.usbmodem1101, /dev/ttyACM0, COM3).
buddy/scripts/)bash# Push a subset of files over USB-serial python3 buddy/scripts/push.py --port <PORT> --files apps/snake.py # Watch device logs python3 buddy/scripts/tail_serial.py --port <PORT> # One-shot REPL exec python3 buddy/scripts/repl_run.py --port <PORT> --script "import os; print(os.listdir('/flash'))"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 12,093 | 4,285 | -65% | 1 | 1 | 0% | 2,184 | 1,248 | -43% | 0 | 0 | — |
case-02 | fail→pass | 9,541 | 2,923 | -69% | 1 | 1 | 0% | 1,680 | 872 | -48% | 0 | 0 | — |
case-03 | fail→pass | 6,921 | 2,369 | -66% | 1 | 1 | 0% | 1,368 | 891 | -35% | 0 | 0 | — |
case-04 | fail→pass | 14,043 | 1,639 | -88% | 1 | 1 | 0% | 2,460 | 725 | -71% | 0 | 0 | — |
case-05 | fail→pass | 12,931 | 1,963 | -85% | 1 | 1 | 0% | 2,366 | 734 | -69% | 0 | 0 | — |
case-06 | pass→pass | 5,069 | 1,997 | -61% | 1 | 1 | 0% | 821 | 647 | -21% | 0 | 0 | — |
case-07 | pass→pass | 10,431 | 1,168 | -89% | 1 | 1 | 0% | 1,915 | 599 | -69% | 0 | 0 | — |
case-08 | fail→pass | 8,561 | 2,274 | -73% | 1 | 1 | 0% | 1,604 | 825 | -49% | 0 | 0 | — |
case-09 | fail→pass | 10,523 | 1,720 | -84% | 1 | 1 | 0% | 1,841 | 697 | -62% | 0 | 0 | — |
case-10 | fail→pass | 10,344 | 1,339 | -87% | 1 | 1 | 0% | 1,668 | 660 | -60% | 0 | 0 | — |
case-11 | fail→pass | 9,250 | 1,469 | -84% | 1 | 1 | 0% | 1,647 | 657 | -60% | 0 | 0 | — |
case-12 | fail→pass | 9,546 | 2,089 | -78% | 1 | 1 | 0% | 1,974 | 820 | -58% | 0 | 0 | — |
case-13 | fail→pass | 14,349 | 2,922 | -80% | 1 | 1 | 0% | 2,753 | 887 | -68% | 0 | 0 | — |
case-14 | fail→pass | 11,116 | 1,672 | -85% | 1 | 1 | 0% | 2,060 | 707 | -66% | 0 | 0 | — |
case-15 | fail→pass | 5,164 | 1,624 | -69% | 1 | 1 | 0% | 1,035 | 615 | -41% | 0 | 0 | — |
case-16 | fail→pass | 4,591 | 1,928 | -58% | 1 | 1 | 0% | 821 | 775 | -6% | 0 | 0 | — |
case-17 | fail→pass | 9,413 | 1,825 | -81% | 1 | 1 | 0% | 1,787 | 713 | -60% | 0 | 0 | — |
case-18 | fail→pass | 9,071 | 2,993 | -67% | 1 | 1 | 0% | 1,801 | 860 | -52% | 0 | 0 | — |
case-19 | fail→pass | 5,256 | 1,376 | -74% | 1 | 1 | 0% | 1,080 | 626 | -42% | 0 | 0 | — |
case-20 | pass→pass | 9,394 | 5,568 | -41% | 1 | 1 | 0% | 1,969 | 1,701 | -14% | 0 | 0 | — |
case-21 | pass→pass | 25,108 | 18,650 | -26% | 1 | 1 | 0% | 4,901 | 4,907 | +0% | 0 | 0 | — |
case-22 | pass→pass | 18,067 | 11,124 | -38% | 1 | 1 | 0% | 3,888 | 2,649 | -32% | 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 +77 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.