docs(gap-baseline): record OpenCode Go as a confirmed, deferred gap

docs/product-technical-gap-baseline.md

@@ -2627,3 +2627,17 @@ Higgins, S. S., Crepalde, N., & Fernandes, L. (2021). Segmented multiplexity: A
 **900-second clarification.** The historical `NoemaRepairDeadlineExceeded` from the html4tree incident came from the retired caller repair path. The three literal `timeout --kill-after=20 900` invocations still present in `opencode-review-dispatch.yml` are separate containment limits for untrusted test-measurement commands; they are not model or Noema inference timeouts. Telemetry and runbooks must report the command class and phase separately.
 
 **Evidence / acceptance.** Permanent tests forbid retry/deadline/sampling symbols in the caller and prove one gateway request, one attempt annotation, control-character-safe telemetry, missing-value rejection, valid trailing-comma normalization, and exact changed-line guidance. Fresh exact-head repository checks and reviews remain the admission authority; predecessor-head evidence is not transferable. The remaining runtime work is to preserve distinct `request_too_large`, discovery, rate-limit, provider transport, malformed-output, stale-head, and sandbox-command-timeout categories in hosted logs.
+
+## OpenCode Go is not a registered provider source — confirmed gap, deferred implementation (2026-09-02)
+
+**Status:** Confirmed gap, not yet implemented. Recorded here per this session's throttle agreement (org-wide GitHub Actions capacity at the 60-concurrent-runner plan ceiling) — implementation deferred to a later, non-urgent iteration rather than opened as a new PR immediately.
+
+**Question investigated.** Whether OpenCode Go shares credentials/discovery with the already-registered `opencode_zen` provider source in `contextual-orchestrator`, or needs its own.
+
+**Evidence, independently verified from two sources:**
+1. `contextual_orchestrator/model_discovery.py` on `contextual-orchestrator`'s protected `main` (fetched directly, not via code search) registers `opencode_zen` (`credential_name="OPENCODE_ZEN_API_KEY"`) as a `ProviderModelSource`. No `opencode_go`, `OPENCODE_GO`, or "OpenCode Go" string exists anywhere in the repository (verified by direct grep of the fetched source; a code-search cross-check returned the same zero count).
+2. OpenCode's own documentation (`opencode.ai/docs/go`, fetched directly) states Go requires signing in to Zen, then *subscribing to Go separately* to obtain a distinct API key ("You sign in to OpenCode Zen, subscribe to Go, and copy your API key") — a paid ($10/month) subscription, not an automatic Zen entitlement. Go's endpoints are also distinct: `https://opencode.ai/zen/go/v1/{responses,chat/completions,messages}`, versus Zen's own `https://opencode.ai/zen/v1/...` paths, and OpenCode configs address Go models with a distinct `opencode-go/<model-id>` identifier.
+
+**Conclusion.** This is a genuine gap, not a naming/discovery oversight: OpenCode Go is a separate paid product requiring its own credential and its own `ProviderModelSource` registration (own `credential_name`, e.g. `OPENCODE_GO_API_KEY`, and the `/zen/go/v1/...` base URL) — it cannot be reached through the existing `opencode_zen` registration regardless of what's configured. No `.env`/secret currently exists for it either.
+
+**Next step (deferred, not urgent).** Add a `ProviderModelSource` for `opencode_go` in `contextual_orchestrator/model_discovery.py` following the same pattern as `opencode_zen`, gated on an `OPENCODE_GO_API_KEY` secret being present (absent-by-default, matching this repo's existing fail-closed provider-registration convention). Requires a Go subscription and its API key to actually exercise in CI; until then this stays a documented, deferred gap rather than a blocking one.