docs(evaluator): add phase5 post-e3 audit handoff

[briansrls]

Summary

• add a docs/audit Phase 5 post-E3 closure handoff for assignment Evaluator Phase 5 post-E3 closure audit handoff #1904
• anchor valiant-carp-10 #1857 as the closed E6-G1.a Option 3 mechanism demonstration authorized by docs(evaluator): add E6 G1a option3 worker brief #1853
• keep Q-Reification / ReflectedProgram, E2 Descent proof consumer, G1.b X1.b S1/S3 generic dispatch, and SymbolicCost runner predicate work explicitly open
• name the next audit queue, triggers, and STOP conditions without touching evaluator Rust/substrate/runner code

Validation

• git diff --cached --check -- docs/audit/r3-evaluator-phase5-post-e3-closure-handoff.md

Closes #1904

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 1a2ed9ad · Trigger: schedule
• Comparison: origin/main @ 54419bad ... review/pr-1905-1a2ed9ad @ 1a2ed9ad
• Thinking: 18s wall

---

Verdict: APPROVE — The only change is a new audit handoff doc (docs/audit/r3-evaluator-phase5-post-e3-closure-handoff.md). Nothing here touches substrate, compiler Rust, tests, or CI. Against the rubric: no modeling-substrate concerns apply; CODING.md / TESTING.md target code and tests and are untouched; INVARIANTS.md / modeling-discipline.md are satisfied in spirit—the packet scopes what #1857 did and did not close, lists residuals with named briefs/carriers, and gives explicit STOP conditions so audits stay fail-closed on authority claims rather than inventing readiness. No concrete violations tied to specific diff lines.

Exploratory (optional): Spot-checked that paths cited in the new doc (e.g. e6_g1a_option3_static_lens_test.rs, phase4 handoff, G1.b brief) exist in this tree; nothing stands out as inconsistent for reviewers using this packet.

[briansrls]

Review metadata

• Provider / model: codex / codex-default
• Commit: 87c20098 · Trigger: schedule
• Comparison: origin/main @ 8ca25298 ... review/pr-1905-87c20098 @ 87c20098
• Thinking: 29s wall

---

Verdict: APPROVE. This looks clean, no concerns.

The diff is a docs-only audit handoff, and it stays precise about scope: it explicitly limits #1857 to the narrow Option 3 mechanism demo and keeps the residual work, triggers, and STOP conditions named in-repo rather than inventing new authority. I did not see any concrete violation of INVARIANTS.md, docs/modeling-discipline.md, CODING.md, or TESTING.md in this diff.