[SCITT] Escalation and hold-window semantics — a companion layer for draft-munoz-scitt-permit-profile

["David S. Rose" <[e-mail]>]

Hi all,

I'm David S. Rose. I build and operate a governed personal-AI-agent
deployment, and I've been following the agent-drafts discussion here
with interest.

The permit profile (draft-munoz-scitt-permit-profile-01) scopes itself
to the pre-execution decision record, and — as far as I can see —
escalation workflows, hold windows, and monitor constructs appear
nowhere in it: "challenge" is a decision value whose downstream
semantics the document leaves open. The CHAP discussion in August
was circling the same ground from the human-authority side.

I have been operating a construction for exactly that layer and would
like to contribute its semantics:

- Two escalation shapes with opposite timeout defaults: an approver
  gate that fails closed (expiry -> deny), and a veto window that
  releases on expiry only when an independent monitor's attested
  last-processed stream position shows it actually consumed the
  proposal before the window closed — silence as no-objection only
  when the objector was demonstrably reading — with degradation to
  gate semantics otherwise.

- An adjudication vocabulary that records an approval-over-objection
  as an override, rather than an ordinary approve.

- A deliberate distinction between "awaiting an approver" and "staged
  pending conditions" — different pauses with different safe defaults.

A short standalone specification of the release interlock ("Attested
Staged Release", v0.1 draft) is attached for discussion. Offered for
incorporation into the permit-profile family or as a companion
profile, whichever the group prefers.

-David