Run the host tests and compile the affected ESP example before submitting a change; commands and toolchain versions are in README.
Keep protocol validation, profile interpretation, BLE lifecycle and application policy separate. Unknown values remain raw or explicitly experimental. Include tests for rejection/failure paths and preserve existing numeric API IDs.
Parameter confirmations should identify the marketed model, firmware version, register/table, proposed scale and a corresponding display/app value. Do not publish serial numbers, logger PNs, site information, credentials, complete phone bug reports, private captures or decompiled proprietary code. Use synthetic/minimized fixtures for public regression tests.
Research tools and field-laboratory firmware belong outside this library. The runtime, builds and tests must work without sibling research directories.
Do not add undocumented register writes or inverter-control experiments. Control support requires a separate safety and recovery design.