m5-onboard

End-to-end onboarding for a freshly-plugged-in M5Stack ESP32 device (Cardputer, Cardputer-Adv, Core, CoreS3, Stick) — detect on USB, flash UIFlow 2.0 firmware, and install the Claude Buddy MicroPython app bundle. Use whenever the user plugs in or wants to flash/provision/reset an M5Stack or ESP32 bo

By anthropics · 1,978 installs

npx skills add anthropics/claude-plugins-official --skill m5-onboard

Source repository · Upstream listing

M5Stack Onboarding This skill automates the full cold start workflow for an M5Stack ESP32 device: detect on USB, identify model, flash UIFlow 2.0, and push a MicroPython app bundle onto /flash/ so the device boots into user software. The apps we ship (Claude Buddy, Snake, Hello) talk over BLE or USB. The workflow runs on macOS, Linux, and Windows; the skill was developed against an M5Stack Basic v2.6 (CH9102 bridge, ESP32 D0WDQ6 V3, 16 MB flash) and generalized to cover the rest of the Core family, with the Cardputer Adv (ESP32 S3, native USB) as the current default target. Where the scripts live This skill ships as part of the cwc makers plugin for reference, but the executable scripts and the buddy/ app bundle live in a local clone of https://github.com/moremas/build with claude (the /maker setup command creates this clone). Run every scripts/ .py invocation below from inside that clone's onboard/ directory so apps buddy resolves to the sibling buddy/device/ payload. When to use Use this when a user plugs in an M5Stack device and wants it provisioned. The decision tree: Fresh/unknown device → run onboard.py apps buddy end to end (detect → identify → flash → install apps). This is the default path. Already flashed device, user just wants apps installed/refreshed → run install apps.py src buddy (or any src <path to a directory of .py files). Flashed device, something feels broken → run smoke test.py (I2C + LCD + speaker + button check). User wants to know what's on the bus / what the device can do → smoke test.py . If multiple devices are plugged in, ask which port to target — don't guess. If the user is provisioning a device they previously worked with (e.g. "same thing as last time" or "another Buddy"), default to apps buddy unless they say otherwise. Which variant to assume The rig this skill lives on provisions Cardputer Adv boards overwhelmingly, so onboard.py now defaults to variant cardputer adv . In practice that means: If the user says nothing about the model, go with the default. They're almost certainly holding a Cardputer Adv. If the user says "Cardputer" (no "Adv"), ask — the two models share a form factor but take different firmware images, and flashing the wrong one boot loops the device. If the user names any other board ("Core2", "CoreS3", "Basic", "Fire"), pass the matching variant explicitly — the default won't apply. The chip is ESP32 S3 either way, and detect.py won't be able to tell Cardputer from Cardputer Adv before UIFlow is flashed (same native USB JTAG VID, no pre flash I2C probe). So this is a user intent question, not a hardware fingerprint one. The workflow The main orchestrator is scripts/onboard.py . It drives the sub scripts in order and handles the handoffs between them (waiting for reboots, capturing MAC, reporting progress). Prefer calling it directly over stitching the sub scripts yourself unless the user asks for a partial run. The default provisioning command (fresh Cardputer Adv, install the buddy bundle): How to invoke this from Claude Code's Bash tool. Do NOT call onboard.py as a foreground Bash command. The Bash tool captures output and does not stream it back to the assistant until the command exits — and this command runs 2–3 minutes. That silence looks identical to a hang, and the assistant will usually give up before the button dance prompt ever reaches the user. Instead, always run with run in background: true , tee to a log file, and then use the Monitor tool (or periodic tail via Read) to surface stage banners, heartbeats, and prompts to the user in real time. 2 &1 is not the fix — all progress already writes to stderr, which a terminal shows fine. The fix is streaming semantics, not redirection. The pattern that works: Relaying physical steps to the user (REQUIRED) The flash stage cannot proceed without a manual button press on native USB boards — there is no software path. When the monitored log shows Enter download mode (or the script appears to wait at the FLASH stage), you MUST stop and tell the user to do the following on the back of the Cardputer , in your own words, before continuing: 1. Press and hold the G0 button 2. While still holding G0, briefly press and release the RST button 3. Keep holding G0 for about one more second, then release it 4. The screen should go fully dark — that means download mode is active If the device reboots into UIFlow instead of going dark, tell the user G0 was released too early and to try again holding it longer. Do not move on, retry the script, or attempt a software workaround until the user confirms the screen is dark — the flash will not start otherwise. The same applies to any later Manual reset prompt: relay the physical step and wait for the user. Users running onboard.py directly in their own terminal (not via Claude Code) will see all output live — no changes needed there. If port is omitted, detect.py picks the most likely candidate across all three OSes: native USB ESP32 S3 ( /dev/cu.usbmodem on macOS, /dev/ttyACM on Linux, COMx on Windows), or a CH9102/CP210x UART bridge on older boards. Bluetooth serial ports are filtered out. If multiple candidates are present, it asks. The known apps name buddy resolves to the buddy/device/ directory in this repo (custom launcher + Hello + Claude Buddy BLE client + Snake). Any other apps value is treated as a filesystem path. To skip re flashing and just push (or refresh) the apps onto an already provisioned device: Where <PORT is whatever detect.py printed on the last full run — for example /dev/cu.usbmodem1101 , /dev/ttyACM0 , or COM3 . Stages 1. Detect ( detect.py ) — enumerate serial ports, filter to USB UART bridges (CH9102 vendor 0x1A86 , Silabs CP210x 0x10C4 , FTDI 0x0403 ) or the ESP32 S3 native USB JTAG interface ( 0x303A ). Probe with esptool to confirm the chip. Port names differ per OS ( /dev/cu.usbmodem on macOS, /dev/ttyACM / ttyUSB on Linux, COMx on Windows) but pyserial abstracts that. 2. Identify ( detect.py ) — alongside port discovery, detect.py reads the factory test partition signature and/or scans I2C once UIFlow is on, and cross references references/hardware signatures.md to suggest the right firmware variant (Basic 16MB, Core2, CoreS3, Cardputer Adv, etc.). User facing variant choice happens via onboard.py variant ; there is no separate detect.py identify flag. 3. Fetch firmware ( fetch firmware.py ) — query the M5Burner manifest API and download the appropriate UIFlow 2.0 binary into the system temp dir. Cached between runs — safe to clear the cache anytime, it just re downloads. 4. Flash ( flash.py ) — esptool write flash 0x0 <image at 460800 baud for UART bridges, no stub at 115200 baud for native USB S3 devices. 921600 fails intermittently on the CH9102 bridge — do not increase it. Native USB flash can intermittently throw Lost connection, retrying mid erase; esptool recovers. The post flash watchdog reset teardown step can fail even when the flash itself succeeded — flash.py parses esptool's stdout, treats that specific failure pattern as non fatal when Hash of data verified appeared, and onboard.py falls back to flash.native reset() and then manual RESET coaching if needed. 5. Install apps (optional, install apps.py ) — paste mode REPL upload of every .py from a source directory into /flash/ , then reboot via repl reset (DTR/RTS is a no op on native USB — don't reach for it). Source layout: root .py → /flash/ , apps/ .py → /flash/apps/ (UIFlow's stock launcher scans that). When the bundle ships a root main.py , install apps.py also sets NVS boot option=2 so UIFlow's own launcher doesn't run and our main.py takes over the boot flow — critical for BLE using apps on ESP32 S3 (see gotchas below). 6. Smoke test (optional, smoke test.py ) — I2C scan, LCD test pattern, speaker beep, button read. Critical gotchas (baked into the scripts — do not second guess) These are things the scripts already handle correctly but which you should not override if the user asks you to "just run esptool manually" or similar: Native USB ESP32 S3 boards (Cardputer, Cardputer Adv, CoreS3) require a physical BtnG0+BtnRST dance to enter download mode. There is no software path. The chip has no DTR/RTS bridge, so nothing esptool or pyserial can do will put it into the ROM bootloader — the user has to hold GPIO0 low across a reset pulse with the hardware buttons. On Cardputer Adv specifically both buttons (BtnG0 and BtnRST) are on the back of the device — small, flush mounted, often easiest to press with a fingernail. onboard.py: wait for download port prompts for this at runtime during FLASH: press and HOLD BtnG0, briefly press BtnRST, release BtnRST first, keep holding BtnG0 for ~1 more second, release BtnG0, screen should be fully dark. If the device reboots back into UIFlow instead, BtnG0 was released too early — the coaching retries and tells the user to hold it longer. Do NOT try to automate this with esptool before default reset or pyserial's DTR/RTS; both are no ops on native USB (the pins aren't wired to EN), and adding them just hides the real prompt. Do not unplug the device during FLASH. Especially on native USB. A mid flash disconnect leaves the internal flash in an inconsistent state. Mask ROM is usually reachable afterwards (press BtnG0 alone on the back, or do the full BtnG0+BtnRST dance), so the recovery is just to re run m5 onboard go — it's idempotent and will re enter download mode, re flash, re push apps. Don't panic and don't start opening the case; the mask ROM is in silicon and survives a corrupted flash as long as the USB PHY is intact. Baud rate is 460800 on UART bridges, 115200 with no stub on native USB. Not 921600 on either. The CH9102 bridge loses sync on erase flash at 921600 (not theoretical — it fails). Native USB's stub baud bump path produces "Lost connection" mid flash; 115200 no stub is counterintuitively faster end to end because it never fails. NVS writes must use set str , not set blob (relevant to install apps.py 's boot option setter). UIFlow's startup calls nvs.get str() and ESP IDF tags blob and string entries separately. A blob tagged key returns ESP ERR NVS NOT FOUND to get str , and the device boot loops. If a prior attempt wrote a blob, call nvs.erase key(name) before set str . REPL multi line blocks need paste mode. Sending try: / except: line by line makes the REPL accumulate indentation forever. Use Ctrl E to enter paste mode, send the block, Ctrl D to execute. mpy repl.py wraps this. Hard reset is DTR=False, RTS=True, 100ms, RTS=False — but only on UART bridge devices. On native USB ESP32 S3 boards the DTR/RTS lines aren't wired to EN/GPIO0, so that pulse is a silent no op. Use mpy repl.repl reset() (sends machine.reset() through the REPL) for post install reboots on those devices — install apps.py already does this. If you bypass install apps.py and stitch your own flow, don't reach for DTR/RTS on a usbmodem port and expect a reboot; files will be on disk but the old code will still be running. That regression bit us once. The idle heap debug loop is normal. UIFlow 2.0 prints asyncio diagnostics while waiting at the pairing screen. Don't interpret it as a hang. Cardputer Adv (ESP32 S3) BLE peripherals require NVS boot option=2 + a custom main.py . UIFlow's default boot option=1 starts a background Flow pairing BLE advertise that wedges the NimBLE controller — subsequent gap advertise(adv data=...) calls from user code hit OSError( 519) "Memory Capacity Exceeded" regardless of payload shape, and the device ends up advertising with empty AD fields that iOS and the desktop Claude Buddy app filter out. The bundle's main.py lives at /flash/ and takes over the boot flow (sh