fix(mcp): exempt the reviewed discovery allowlist from the mcp:false veto

The signal registry's `mcp:false` flag vetoes a metric-status resolution
so an unreviewed non-exposed signal cannot leak through the id/display-name
path. The metric-status discovery allowlist is a deliberate, human-named
exposure — co-equal to `mcp:true` — so its ids must stay resolvable, or
compare_metric / get_metric_baseline would advertise a metric in discovery
that then refuses to resolve. Exempt the allowlist from the veto; the guard
still fires for genuinely-unreviewed mcp:false ids (workout HR stats).

src/lib/mcp/__tests__/rich-reads.test.ts

@@ -133,16 +133,35 @@ describe("resolveRichMetric", () => {
   // v1.30.4 (C2) — the signal registry's `surfaces.mcp` flag is documented as
   // the SINGLE source of truth for MCP exposure, but `resolveRichMetric`'s
   // metric-status-registry path (exact id / display-name match) used to
-  // resolve an id regardless of that flag. `CARDIO_RECOVERY` /
-  // `WRIST_TEMPERATURE` / `SLEEP_SCORE` are all valid `METRIC_STATUS_IDS`
-  // entries but carry `mcp:false` in the signal registry — pin that they
-  // now stay unreachable, closing the contract gap (no PHI actually leaked
-  // through it, but a future `mcp:false` signal without this guard would).
+  // resolve an id regardless of that flag. `AVERAGE_HEART_RATE` /
+  // `MAX_HEART_RATE` are valid `METRIC_STATUS_IDS` entries but carry
+  // `mcp:false` in the signal registry AND are NOT on the reviewed
+  // metric-status discovery allowlist — pin that they stay unreachable,
+  // closing the contract gap (no PHI actually leaked through it, but a future
+  // `mcp:false` signal without this guard would). The reviewed discovery ids
+  // (`CARDIO_RECOVERY` / `WRIST_TEMPERATURE` / `SLEEP_SCORE` / …) are the one
+  // deliberate exemption and stay resolvable — asserted below.
   it("honours the signal registry's mcp:false as a hard veto over a metric-status hit", () => {
-    expect(resolveRichMetric("CARDIO_RECOVERY")).toBeNull();
-    expect(resolveRichMetric("cardio recovery")).toBeNull();
-    expect(resolveRichMetric("WRIST_TEMPERATURE")).toBeNull();
-    expect(resolveRichMetric("SLEEP_SCORE")).toBeNull();
+    expect(resolveRichMetric("AVERAGE_HEART_RATE")).toBeNull();
+    expect(resolveRichMetric("MAX_HEART_RATE")).toBeNull();
+    expect(resolveRichMetric("max heart rate")).toBeNull();
+  });
+
+  // v1.30.4 (coverage review) — the reviewed metric-status discovery allowlist
+  // is a deliberate MCP exposure, co-equal to a signal `mcp:true`. A
+  // `mcp:false` on the same signal must NOT strip these ids' resolvability, or
+  // `compare_metric` / `get_metric_baseline` would advertise a metric in
+  // discovery that then refuses to resolve.
+  it("exempts the reviewed metric-status discovery allowlist from the mcp:false veto", () => {
+    expect(resolveRichMetric("CARDIO_RECOVERY")?.measurementType).toBe(
+      "CARDIO_RECOVERY",
+    );
+    expect(resolveRichMetric("WRIST_TEMPERATURE")?.measurementType).toBe(
+      "WRIST_TEMPERATURE",
+    );
+    expect(resolveRichMetric("SLEEP_SCORE")?.measurementType).toBe(
+      "SLEEP_SCORE",
+    );
   });
 
   it("still resolves a metric-status id the signal registry does NOT mark mcp:false", () => {

src/lib/mcp/rich-reads.ts

@@ -214,9 +214,19 @@ function fromRegistry(id: string): RichMetric | null {
   // the SAME id in the signal registry as a hard veto over a metric-status
   // hit; an id absent from the signal registry (most of `METRIC_STATUS_IDS`)
   // is unaffected — metric-status stays authoritative when the two don't
-  // disagree.
+  // disagree. The one exemption is the reviewed metric-status discovery
+  // allowlist (`REVIEWED_DISCOVERY_IDS`): those ids are a deliberate,
+  // human-named MCP exposure — co-equal to a signal `mcp:true` — so a
+  // `mcp:false` on the same signal must NOT strip their resolvability, or
+  // `compare_metric` / `get_metric_baseline` would advertise a metric in
+  // discovery that then refuses to resolve.
   const sig = getSignal(id);
-  if (sig && sig.kind === "measurement" && sig.surfaces.mcp === false) {
+  if (
+    sig &&
+    sig.kind === "measurement" &&
+    sig.surfaces.mcp === false &&
+    !REVIEWED_DISCOVERY_IDS.has(id)
+  ) {
     return null;
   }
   const metric: RichMetric = {
@@ -400,6 +410,21 @@ const METRIC_STATUS_DISCOVERY_IDS = [
   "ENERGY_EXPENDITURE_KJ",
 ] as const satisfies readonly MetricStatusMetricId[];
 
+/**
+ * The reviewed metric-status discovery allowlist as a membership Set. Each
+ * member is a DELIBERATE, human-named MCP exposure (v1.30.4 coverage review) —
+ * co-equal to a signal-registry `mcp:true`, just carried in the metric-status
+ * layer instead of the signal layer. `fromRegistry`'s `mcp:false` veto exempts
+ * these ids so a reviewed metric stays resolvable (`compare_metric` /
+ * `get_metric_baseline` / `detect_changepoints`); the veto still fires for any
+ * UNREVIEWED `mcp:false` signal that happens to carry a metric-status entry
+ * (the two mental-health screeners and every `ENV_*` signal are absent from
+ * this list and have no metric-status entry, so they stay unreachable).
+ */
+const REVIEWED_DISCOVERY_IDS: ReadonlySet<string> = new Set(
+  METRIC_STATUS_DISCOVERY_IDS,
+);
+
 /**
  * The metric-status-only ids `list_metrics` / the inventory resource /
  * `search` can now discover: the registry id (also the `metric:` id `fetch`