carlos / platform Public

Astra stage B round 3 brief

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Paul Campbell pushed by claude 2 weeks ago 919662d246d3cbc5c9848b26c2162741bc8a3271 parent b897dbb

1 file changed, +21 −0

docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-3.md +21 −0

CI failed — run details

diff --git a/docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-3.md b/docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-3.md

new file mode 100644

index 000000000..c57b9959d

--- /dev/null

+++ b/docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-3.md

@@ -0,0 +1,21 @@

+# Astra round 3 brief: region HA stage B plan rev 3

+**Review:** `docs/superpowers/plans/2026-10-10-region-ha-1b-next-stage-b.md` rev 3 (b897dbb7). The "Rev 3: round 2 dispositions" table answers your round 2 (`CODEX-PLAN-REGION-HA-STAGE-B-R2.md`): 9 accepted, 2 changed. Briefs 1 and 2 still apply.

+**What changed:**

+- **C7 is withdrawn.** Completion journals the probe stop's receipt reference on the serving op. The first serving wake verifies it and hands it to `ColdHandover.Run` as the exact prior, which is your round 2 suggestion; named change N6.

+- **New claim ask C10**, an activation gate: a CC9 discharge of the participation by the completed probe evidence.

+- **New claim ask C11:** root pins `install_stamp` only for P's opening. Until it lands, the fallback is one launchable attempt per target.

+- **O2 → O2'.** There is a three-source key-evidence check: `instance.key`; the platform key plus an adoption record; or the config key. It is defined in the instance-key plan, which is still being revised on branch `instance-key-build`.

+- **The 200-seed lifecycle scenarios** are ported to `productionClaimLocal`.

+**Judge:**

+1. Is each round 2 finding closed?

+2. Is N6 sound against `ColdHandover.Run` and root's receipt verification? Can the serving wake accept a receipt that doesn't belong to this host's last writer?

+3. Is C10's discharge rule safe? Above all: can a lease loss during the probe leave a participation discharged while work is still running?

+4. Is C11's fallback safe and bounded?

+5. Is the O2' check correct, and does the negative case refuse? Is the pinner interface stable enough, given the adoption record's shape isn't final?

+6. Are the ported lifecycle scenarios real coverage?

+7. Anything new?

+Give findings P1/P2/P3 with file:line, then a verdict: ready to build, or not.