KB-6BAC

Macro-3 Channel Final Recommendation — R2-B2 (2026-06-19)

4 min read Revision 1
laws-newR2-B2macro-3channelfinal-recommendationrecommendation-onlynon-authorizingowner-gated2026-06-19

Macro-3 Channel Final Recommendation — R2-B2 (2026-06-19)

Date: 2026-06-19 · Workstream: R2-B2-MACRO-3-OPTION-D-OWNER-DELEGATE-DECISION-PACKAGE-2026-06-19 (Deliverable 6 of 60) · Editorial revision: rev1 Class: channel final recommendation · READ-ONLY · NON-ENACTING · NON-AUTHORIZING · RECOMMENDATION_ONLY — NOT AUTHORITY · CHANNEL_AUTHORITY_DRIFT guard · NO channel selected · NO write performed.

Metadata convention. Editorial revision (rev1) only. Storage revision/content_length authoritative at read time.

Channel-authority lock. This is a recommendation, not a selection. It wires nothing, installs nothing, promotes no contract. RECOMMENDATION_ONLY — NOT AUTHORITY — OWNER_GATE_REQUIRED — FUTURE_TECHNICAL_DESIGN_REQUIRED. Any wording that reads as a selection is CHANNEL_AUTHORITY_DRIFT → HOLD. The channel is an internal of B2 (B2-AC-7) and is replaceable.


0. Status and non-authorization

STATUS: PASS — engineering / decision-prep. Carries Macro-2's channel ballot one step into a final recommendation, still selecting nothing. Engineering PASS ≠ authority PASS. Default: HOLD.

1. Purpose

Answer macro question 1 — which channel should GPT recommend to Owner: host cron or agent-api? — as a final, conditional recommendation.

2. Sources / evidence read

Macro-2 channel decision ballot, host-cron + agent-api decision memos, weighted comparison, proof-gap closure, channel non-authority + rejected-channel registers; FRESH 2026-06-19 (list_docker, query_pg). Main process, no reader-agents.

3. Accepted baseline (carried)

Macro-2 ballot: host cron + agent-api are the only candidates (neither selected); pg_cron / job_queue = risky/future; manual one-shot = reject-as-standing (that is B5). Channel sits inside B2 and is swappable.

4. Evidence / analysis (FRESH 2026-06-19)

  • host cronwf_host_crontab_snapshot = 54 jobs, 0 birth; the channel is readable and unwired; lowest blast radius; proven as the sibling-scanner channel.
  • agent-api executor — container Up (healthy) :8090; dot_agent_api_contract = 2 (KG-EXPLAIN), 0 birth-bound; runtime switches OFF (execute_enabled/real_run_enabled=false, dry_run_only=true). Governance-aligned (contract-bound, observable) but not yet wired to birth and would need a DRY_RUN-gated contract promotion (Owner-gated).

5. Recommendation / matrix / result (RECOMMENDATION_ONLY — NOT AUTHORITY)

Owner priority Recommended channel Why
observability / contract-bound promotion / auditability agent-api executor governance traceability; observable; contract-bound
simplicity / replaceability / faster pilot / lowest blast radius host cron unwired, smallest surface, trivial rollback (remove crontab line)
forced single pick for Macro-4 readiness lean agent-api governance traceability — but Owner approval still required

Nothing is selected. The default disposition is HOLD until the Owner decides (Deliverable 15 poses the question; Deliverable 48 ballots it).

6. Owner-gated future work

Selecting the channel, wiring host cron, or binding/promoting the agent-api contract are Owner acts via Điều 37→32; forbidden now.

7. What remains unresolved

The channel is not selected; GATE-4 (channel) remains open and non-waivable.

8. Ready for GPT/Codex review

Yes — Codex should attack the "forced single pick = agent-api" lean and confirm it is not phrased as a selection.

Back to Knowledge Hub knowledge/dev/laws-new/newlaws/consolidation/macro3-channel-final-recommendation-r2-b2-2026-06-19.md