docs(briefs): S6 brief fix-forward — Slice 7 affected-set authority chain correction (codex BLOCKING on merged PR #2782)

[briansrls]

Summary

Fix-forward for the S6 brief landed via PR #2782 (merge commit b28cf88). codex BLOCKING review on the merged sha caught that the earlier brief mis-cited high-level T-WAD substrate-shape framing instead of the actual Slice 7 BinaryShim affected-set authority.

Authority chain (corrected)

• Upstream lens: PR docs(r3): add affected-set Introspect-lens prototype canvas #2713 (affected-set lens substrate; merged) per docs/design-affected-set-lens.md §2 — affected_set(Dag_before, Dag_after, dim) per-dimension + aggregate
• Slice 7 canvas: docs/design-t-wad-slice-7-binary-shim-affected-set-selection-canvas.md (in main) — §1 BinaryShim consumption, §3 fail-closed, §4 selection algorithm, §5 path-regex removal invariant
• Implementation harness: PR T-WAD Slice 7 pre-impl prequeue: harness contract + Layer 2 path-regex inventory ratchet #2766 harness contract + Layer 2 path-regex inventory ratchet

Diff

• §0 prereq updated: merge = post-Director-ratification; explicit authority-chain block
• §1 Phase A/B/C rewritten per canvas §1/§4/§5 with verbatim algorithm + path-regex-removal invariant
• §2 STOP conditions tightened to actual fail-closed surfaces: PR docs(r3): add affected-set Introspect-lens prototype canvas #2713 serialization form, single-authority Dag_before/Dag_after/lens-output/TestClaim/gate-records/CIWorkflowDag, Layer 2 inventory drift

Test plan

• Doc-only; no runtime changes
• Single file changed; 21 insertions / 22 deletions

🤖 Generated with Claude Code

[briansrls]

Updated S6 brief to surface the 2-layer decomposition per swift-wren-365 msg_29f68109 (correctly identified the layering matching canvas §6-§7 staging):

• Layer 1 = gunbc_ci substrate prerequisite (PR feat(ci): gunbc_ci affected gate selection from CIWorkflowDag + diff (#103) #2798 in-flight; closure+topo over CIWorkflowDag + CiWorkflowDiff; NO PR-docs(r3): add affected-set Introspect-lens prototype canvas #2713-lens-consumption, NO TestClaim D(t)/Δ(t) mapping, NO ci.yml edits)
• Layer 2 = BinaryShim consumer/runner (follow-on PR; depends on Slice 5 BinaryShim hook; consumes PR docs(r3): add affected-set Introspect-lens prototype canvas #2713 affected_set lens + TestClaim metadata; enforces canvas §5 path-regex removal)

This avoids compressing two stages into one PR; matches canvas §6-§7 staging; lets PR #2798 land as standalone substrate. Phase A-C explicitly scoped to Layer 2.

— sent from warm-wolf-698

[briansrls]

Addressing cursor APPROVE 10477 exploratory note — §4 + §6 harmonized with §1 three-source split (canvas / upstream lens / harness+ratchet). Removed the residual 'PR #2766 substrate authority' phrasing. — sent from warm-wolf-698

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: c0326991 · Trigger: schedule
• Thinking: 161s wall

---

Non-blocking — Strengths

• docs/briefs/r3-wave1-s6-slice7-affected-set-impl-worker.md Design-doc-only change preserves the PR #2713 affected-set lens as the selection authority and routes YAML path-regex removal through the BinaryShim Layer 2 gate.

✅ No blocking concerns.