fix: close live-run gate and proposal regressions

[Stahl-G]

Summary

Publishes the two fixes proven during the real Codex demo run and closes their finite review findings.

• Accept the first invocation-owned dynamic proposal artifact from revision zero and create its authoritative artifact record atomically.
• Re-evaluate Gate freshness from frozen incident opened_at metadata and target relevance from frozen target_terms.
• Require one producer implementation/version across each Gate batch.
• Add a public runtime regression from Scout proposal acceptance through postcommit verification and the Screener next action.

Exact candidate

• Base: 75db1ecadefcc75a064090dd43b342a4165fb257
• Head: 5931190de91fa13422961ff08dce279e21a56cbb

User impact

Codex no longer fails the first Scout proposal with runtime_envelope_invalid merely because the dynamic artifact has no previous revision. Frozen incident time and target terms no longer produce the two false Gate blocks observed in the demo.

Authority boundary

SQLite ControlStore and receipt-owned revisions remain the sole run authority. Agents remain proposal-only. The fix does not add automatic approval, Gate bypass, finalization, delivery, legacy state, or any new UX workflow.

Validation

• 139 passed, 1 skipped across runtime host, Core v2, next-action, and quality-gate focused suites
• Targeted post-rebase delta: 2 passed
• Ruff: pass
• git diff --check: pass
• One Reviewer plus one targeted delta re-review

Not included

Init authorization, finalized_local, proposal field diagnostics, output-contract enforcement, Gate repair, review HTML, and delivery changes are intentionally reserved for separate RUN-UX successors.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5931190de9

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Refresh stale finalize gate batches

This refresh is only applied in _auditor_action; the terminal finalize path still relies on _has_current_finalize_gate and _apply_finalize_complete, which accept the current finalize report as long as its gates are pass/warning without checking producer_version. In an upgraded live run that already has a v1 finalize gate report, next_action can choose finalize_complete (or remain blocked on a v1 blocking result), so the v2 evaluator fixes for target terms/status incidents are never applied before delivery. Please apply the same evaluator-version refresh to finalize batches before completing finalize.

AGENTS.md reference: AGENTS.md:L104-L104

Useful? React with 👍 / 👎.