Record an independent developer result

Use a separate result directory outside the authenticated package. Copy
RESULT-TEMPLATE.json there and fill it from observations, not expected values.
Identify who operated the trial: human unaided, human with AI assistance, or AI
agent. Record prior familiarity and any help from the implementation authors.
Do not relabel an AI-operated run as an unaided human installation.

Keep the received manifest/signature/public-key identity; archive and inventory
pins; original downloads; exact OS/image/kernel/CPU/filesystem/runtime versions;
fresh account/path choices; all commands, UTC start/end times, stdout/stderr,
exit codes, deviations and assistance. Record the first failing step and preserve
it. A nonzero exit, missing evidence or unresolved outcome is FAIL/INCOMPLETE.
Report separate statuses for installation, isolation, protected operations and
evidence export. Organizational adoption requires separate evidence of actual
organizational use; a passing evaluation or download is not adoption.

On completion or failure, stop new dispatches. Follow documented failure handling
and CLEANUP.txt step 1 to export only after the service is stopped. Preserve the
export SHA-256 and the raw evidence; never restore its state into a live domain.
If safe export cannot complete, preserve existing files and the exact error.
Do not edit the witness, repair a domain, replay setup, silently patch code or
substitute dependencies. Keep a failed first attempt even if a separately
identified future trial is later authorized.

Before any destructive cleanup, verify the export and keep it outside the paths
that cleanup removes. Cleanup --destroy was not exercised in the included Linux
trial. A reviewer should inspect raw state/journals/dispatches and the independent
oracle, not just a screenshot or a PASS string.

The designated reporting destination in historical records is eziah.dev, but a
working public intake was not verified in this gate. Until the publisher provides
an authenticated intake/contact, retain the result locally and use a separately
agreed channel. Do not assume a website form exists or upload private machine
information. Keep an unmodified private original; label any redacted sharing copy
and record its separate digest. No telemetry or automatic submission is required.