Review brief: GeoIP implementation plan 1

You are GPT-6 reviewing an implementation plan independently of its author. Your output is a findings report, not an edit. Do not modify any file.

What to read

1. The plan: docs/superpowers/plans/2026-10-06-platform-geoip-plan1.md.
2. The spec it implements (rev 6, reviewed over five rounds): docs/superpowers/specs/2026-10-06-platform-geoip-design.md. The spec is settled; report where the PLAN fails the spec, not where you would change the spec.
3. docs/superpowers/plans/2026-10-06-platform-geoip-FACTS.md (measured).
4. Every file the plan names, in this worktree, at the lines it cites.

What matters most

Each task is handed to a separate implementer who reads only that task.

• Code that will not compile or will not behave as the plan claims. Wrong signatures, missing imports, fields or helpers that do not exist, a test that would pass for the wrong reason or could never fail, a test that cannot fail before the implementation step.
• Claims about existing code (line numbers, helper names, behaviour) that are wrong. Check each.
• Spec requirements with no task, or a task that implements them differently from the spec (security-relevant ones first: the strip, adoption trust, the door, rejected hops, the tunnel marks, D1 expiry, the publish slot, the verdict).
• Ordering: a task that depends on something a later task produces.
• Concurrency and lifecycle bugs in the code the plan supplies (geoMirror, geoipPublisher, Swappable).

Output

A numbered list, most severe first: severity (P1 = would ship a bug or fail to build, P2 = an implementer would likely get it wrong, P3 = clarity), task and step, evidence with file:line, and a concrete fix. End with a one-line verdict: ready to build, or revise.