Supported recovery workflow after a managed run escalates without a concrete human question #40

I am trying to resolve a human-input escalation while keeping Foreman’s normal safeguards and genuine completion requirement.

With foreman-core 0.4.1, a managed foreman run can stop with ESCALATED and the generic reason semantic assessment requires human input. The saved assessment contains a needs_human value, but no specific question describing the judgment, credentials, clarification or permission required. The active worker is stopped before its review finishes.

The installed CLI exposes hook, run, demo, inspect, runs and extension, with no managed-run answer, approve or resume command. inspect displays the saved result. The hook route appears to manage a separate attached session. TypeSafe’s System One documentation says that the model does not generate reasoning explanations.

What is the supported operator workflow to:

1. Identify the concrete human input that is needed after this generic escalation?
2. Supply that input while preserving the escalated run as a historical result?
3. Continue through a fresh or resumed managed review that still has to satisfy the unchanged verification and completion checks?

Is there an existing state, input or extension mechanism for this workflow that I have missed? If the workflow is not supported yet, which minimal diagnostic information would help investigate it, and would a provider request ID be useful?

I want to retain the human-input safeguard, existing thresholds and genuine FINISHED requirement. This question includes no application source, prompts, request payloads, logs, credentials or account information.