[DRSDavidSoft]

The libavrdude adapter firmware-write path omitted Urclock nometadata even though the CLI path supplies it. On the connected ATmega328P, libavrdude rejected the firmware identity bytes at the end of application flash as overlapping its metadata byte.

Add the matching option for Urclock flash writes, with a regression check that ISP does not receive the option.

Validation: focused programmer tests passed through the stable test runner using installed Go 1.27.1 and an external temporary modfile. The canonical go.mod is unchanged. The pinned Go 1.27.2 download returned HTTP 403, so that exact toolchain remains a CI gate.

This is a draft checkpoint because live coordinated updating has additional unresolved compatibility problems:

• Runtime command options omit configured avrdude_backend/avrdude_adapter, so the primary can select the library despite an explicit CLI preference.
• Pairing the installed adapter with an AVRDUDE 8.0 configuration fails at pickit5_updi configuration parsing. Matching the packaged adapter/configuration passes probing.
• Packaged AVRDUDE 8.3 CLI writes read back bytes 75 72 at reset-vector padding offsets 2 and 3; the current strict verifier rejects those extra differences. The MiniCore 8.0 CLI pair passes the existing full readback proof.

Next action: propagate backend selection through runtime/primary command options with tests, establish the upstream meaning of the AVRDUDE 8.3 padding marker before changing its verifier, then build and validate a coordinated update with the matching package. Do not relax readback verification generically.

The LED investigation uses the supported standalone guarded programming path with the MiniCore 8.0 CLI pair and a persistent host for rainbow streaming. Related: #693, #71, #390. No device backups, settings, or private logs are included.