D050 - The orchestration is scripts in the repository, not steps somebody remembers

Status: decided, late.

Everything in scripts/ was build-side. Running a module in an emulator, pulling the report out of a stream it shares with the emulator's logging, looping until nothing kills the process, harvesting NID pairs from logs - all of it was done by hand, twenty-odd times, and none of it was written down. The analysis tooling was strong and the orchestration existed only in whoever had been doing it.

That is the same failure this project keeps designing against: docs/DECISIONS.md exists because the conversation that produced a decision does not survive, and a remembered procedure is no better.

Three scripts now: run-emulator.ps1, sweep.ps1, harvest-nids.ps1.

They are PowerShell, and that is not a preference. The module is built inside a VM because a Windows mount cannot carry the execute bit (D012), and the emulators are Windows applications. sh for building, PowerShell for running, and the split is where the work is rather than a style choice.

Two environment hazards are recorded in docs/WORKFLOW.md because both cost time and neither is guessable: build/ can end up owned by the VM and unwritable from the host, and Windows PowerShell turns a native command's stderr into a terminating error, so a multipass transfer that succeeds with a warning ends a script.