Macro-4 Codex Review Packet — R2-B2 (2026-06-19)
Macro-4 Codex Review Packet — R2-B2 (2026-06-19)
Date: 2026-06-19 · Workstream: R2-B2-MACRO-4-STAGING-WORKBENCH-IO-CONTRACT-TD-ENTRY-GATE-2026-06-19 (Deliverable 73 of 90) · Editorial revision: rev1
Class: Codex adversarial review packet · READ-ONLY · NON-ENACTING · NON-AUTHORIZING · NO write performed.
Metadata convention. Editorial revision (rev1) only. Storage revision/
content_lengthauthoritative at read time.
0. Status and non-authorization
STATUS: PASS — engineering / design-only. The adversarial brief for Codex on the 90-file package. Engineering PASS ≠ authority PASS. Default: HOLD.
1. Purpose
Tell Codex exactly what to attack before the Owner-delegate accepts Macro-4.
2. Sources / evidence read
All 89 sibling deliverables; Macro-3 Codex packet (AX-1…10) + Codex review; the process caveat (read direct, main process, no reader-agents). Main process, no reader-agents.
3. Accepted baseline (carried)
Macro-3 Codex verdict PASS_WITH_CAVEATS; the process caveat was honored this run (sources read first-hand in the main process; oversized batches decoded via /tmp python = decode-scratch only, never via a reader-agent).
4. Evidence / analysis — attack surface (full list in Deliverable 89)
| # | Attack | Where |
|---|---|---|
| MX-1 | Is any staging deliverable a build artifact (schema/DDL/corpus)? | 6–12, 57–58 |
| MX-2 | Is the IO contract envelope a mega-registry / universal write surface? | 19–28, 68 |
| MX-3 | Is B2's output anything but candidate-only inspect_*? |
30, 35 |
| MX-4 | Is any channel statement a selection / wiring? | 61 |
| MX-5 | Is any S2 statement an assignment / ownership-row write? | 62 |
| MX-6 | Is any Điều 0-G statement adoption / recovery / patch? | 63 |
| MX-7 | Is any bad-input test run / digest produced? | 49–54 |
| MX-8 | Is the entry gate honestly NO-GO; any actual TD present? | 55–56 |
| MX-9 | Is the staging-build gate conflated with the TD gate? | 57–59 |
| MX-10 | Delete-fast / no-production-touch: any production write? | 10–11, 37–48, 81 |
| MX-11 | B5/B7/R1 scope creep? | 66–67 |
| MX-12 | v0.1 overwritten / v0.2 promoted? | 71 |
| MX-13 | Any blocker falsely resolved? | 79 |
| MX-14 | Any deliverable not independently discardable (NOT_LEGO_COMPATIBLE)? | 3 |
5. Contract / requirement / matrix / result
Codex should return a per-MX verdict + any new caveat. Default expectation: MX-1…MX-14 not triggered (self-checked), but Codex is the adversary of record.
6. Owner-gated future work
Owner-delegate acceptance follows Codex review; not enacted here.
7. What remains unresolved
Codex review pending; all blockers OPEN.
8. Ready for GPT/Codex review
Yes — this is the Codex entry point (paired with Deliverable 89's full attack list).