Macro-3 Channel Agent-API Risk Map — R2-B2 (2026-06-19)
Macro-3 Channel Agent-API Risk Map — R2-B2 (2026-06-19)
Date: 2026-06-19 · Workstream: R2-B2-MACRO-3-OPTION-D-OWNER-DELEGATE-DECISION-PACKAGE-2026-06-19 (Deliverable 10 of 60) · Editorial revision: rev1
Class: channel risk map (agent-api) · 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 / decision-prep. The risks the Owner accepts if agent-api is later chosen. Engineering PASS ≠ authority PASS. Default: HOLD.
1. Purpose
State agent-api's downside honestly so the case (D8) and the forced-pick lean are not read in isolation.
2. Sources / evidence read
Macro-2 agent-api memo, weighted comparison, channel proof-gap closure; FRESH switches OFF / 0 birth-bound. Main process, no reader-agents.
3. Accepted baseline (carried)
Governance-aligned ≠ ready; the larger surface is the trade.
4. Evidence / analysis — risk table
| # | Risk | Severity | Note |
|---|---|---|---|
| AA-R1 | Contract promotion scope creep — a birth contract could over-reach beyond inspect-produce | HIGH | B2-only lock; one-column-per-inspector |
| AA-R2 | Switch-flip authority — turning execute_enabled/real_run_enabled on is a powerful, auditable act |
HIGH | must stay Owner-gated; DRY_RUN first |
| AA-R3 | DRY_RUN bypass — a mis-built contract could escape dry-run | MED | proof plan D12 obligation |
| AA-R4 | Larger blast radius than host cron — touches the shared executor service | MED | weigh vs HC-R1 observability gain |
5. Recommendation / matrix / result
If agent-api is chosen, AA-R1…AA-R4 must be mitigated by a DRY_RUN-gated, inspect-only contract proven in staging. Not selected here.
6. Owner-gated future work
Contract design + DRY_RUN proof are Owner-gated (actual B2 TD), not here.
7. What remains unresolved
All four risks remain OPEN; 0 birth-bound contracts exist today.
8. Ready for GPT/Codex review
Yes — Codex should weigh AA-R4 (blast radius) against the observability gain that motivates the lean.