diff --git a/src/review/rule-gate-eval.ts b/src/review/rule-gate-eval.ts
index f6d01e2cd9..975dd65c92 100644
--- a/src/review/rule-gate-eval.ts
+++ b/src/review/rule-gate-eval.ts
@@ -258,3 +258,18 @@ export function rulesBelowClosePrecisionFloor(
 ): BlendedRuleGateEvalRow[] {
   return rows.filter((r) => r.wouldClose >= minDecided && r.weightedClosePrecision != null && r.weightedClosePrecision < floor);
 }
+
+/**
+ * The per-(project, ruleCode) counterpart to {@link rulesBelowClosePrecisionFloor}: same floor/minDecided
+ * check, over an already-fetched {@link RuleGateEvalReport}'s rows instead of the blended ones (no I/O). This
+ * is the "which rule is broken on which repo" view (#8906) — a rule that looks healthy once pooled across
+ * every project can still be systematically wrong on ONE of them, diluted out of the blended figure by every
+ * other project's correct usage of the same rule.
+ */
+export function projectRulesBelowClosePrecisionFloor(
+  rows: readonly RuleGateEvalRow[],
+  floor: number = AUTOTUNE_CLOSE_PRECISION_FLOOR,
+  minDecided: number = AUTOTUNE_MIN_DECIDED,
+): RuleGateEvalRow[] {
+  return rows.filter((r) => r.wouldClose >= minDecided && r.weightedClosePrecision != null && r.weightedClosePrecision < floor);
+}
diff --git a/src/services/operator-dashboard.ts b/src/services/operator-dashboard.ts
index e1e75ae97c..b80b74ad46 100644
--- a/src/services/operator-dashboard.ts
+++ b/src/services/operator-dashboard.ts
@@ -34,7 +34,7 @@ import { computeFleetAnalytics, getFleetHealthSummary, type FleetAnalytics, type
 import { computeAgentHealth, computeCalibration, type AgentHealth, type Calibration } from "../review/ops";
 import { computeGateEval, type GateEvalReport } from "../review/parity";
 import { computeContributorGateEval, contributorFairnessFlags, computeBlendedContributorGateEval, contributorGlobalFairnessFlags } from "../review/contributor-gate-eval";
-import { computeBlendedRuleGateEval, rulesBelowClosePrecisionFloor } from "../review/rule-gate-eval";
+import { computeRuleGateEval, computeBlendedRuleGateEval, rulesBelowClosePrecisionFloor, projectRulesBelowClosePrecisionFloor } from "../review/rule-gate-eval";
 import { computeCycleTimeAggregate, computeFindingAcceptance, type CycleTimeAggregate } from "../review/stats";
 import { loadUpstreamStatus, type UpstreamStatus } from "../upstream/ruleset";
 import { nowIso } from "../utils/json";
@@ -140,6 +140,7 @@ export async function buildOperatorDashboardPayload(
     gateEval,
     contributorGateEval,
     blendedContributorGateEval,
+    ruleGateEval,
     blendedRuleGateEval,
     cycleTime,
     calibration,
@@ -172,6 +173,9 @@ export async function buildOperatorDashboardPayload(
     computeContributorGateEval(env, { days: GATE_ANALYTICS_WINDOW_DAYS, nowMs: Date.now() }),
     // #global-contributor-trust: the SAME data pooled cross-repo into one blended figure per login, same window.
     computeBlendedContributorGateEval(env, { days: GATE_ANALYTICS_WINDOW_DAYS, nowMs: Date.now() }),
+    // #8906: per-(project, ruleCode) gate accuracy -- the "which rule is broken on which repo" view that
+    // computeBlendedRuleGateEval's cross-repo pooling can't provide. Fails safe to an empty report.
+    computeRuleGateEval(env, { days: GATE_ANALYTICS_WINDOW_DAYS, nowMs: Date.now() }),
     // #7984: the SAME review_audit data re-aggregated by RULE CODE (pooled cross-repo) instead of by project —
     // isolates a single systematically-wrong deterministic rule even while its host project's own aggregate
     // still looks healthy. Fails safe to an empty report on any read error.
@@ -214,6 +218,9 @@ export async function buildOperatorDashboardPayload(
   // #7984: rules whose sample has cleared enough volume to trust but whose weighted close precision sits
   // below the SAME floor auto-tune.ts's project-level close-precision breaker uses -- pure fold, no extra I/O.
   const rulesBelowFloor = rulesBelowClosePrecisionFloor(blendedRuleGateEval.rows);
+  // #8906: the SAME floor check, per (project, ruleCode) instead of pooled -- surfaces a rule that's broken on
+  // one specific repo even while the blended figure above still looks healthy.
+  const projectRulesBelowFloor = projectRulesBelowClosePrecisionFloor(ruleGateEval.rows);
   const installedRepos = repositories.filter((repo: RepositoryRecord) => repo.isInstalled).length;
   const registeredRepos = repositories.filter((repo: RepositoryRecord) => repo.isRegistered).length;
   // #1967: map FindingAcceptanceAggregate's field names onto the AcceptanceRateCard's expected shape.
@@ -304,6 +311,17 @@ export async function buildOperatorDashboardPayload(
         value: String(rulesBelowFloor.length),
         delta: rulesBelowFloor.length > 0 ? rulesBelowFloor.map((r) => r.ruleCode).join(", ") : "no rule below floor",
       },
+      {
+        // #8906: the per-(project, ruleCode) counterpart to the blended tile above -- names the specific repo
+        // a rule is broken on, same "name the specific thing" posture (rule codes and repo names are not
+        // personally identifying, unlike the contributor tiles' logins).
+        label: "Rules below close-precision floor (per-project)",
+        value: String(projectRulesBelowFloor.length),
+        delta:
+          projectRulesBelowFloor.length > 0
+            ? projectRulesBelowFloor.map((r) => `${r.project}:${r.ruleCode}`).join(", ")
+            : "no rule below floor",
+      },
     ],
     noiseReduction: [
       {
diff --git a/test/unit/operator-dashboard.test.ts b/test/unit/operator-dashboard.test.ts
index dfe2409cb3..40f840b5c2 100644
--- a/test/unit/operator-dashboard.test.ts
+++ b/test/unit/operator-dashboard.test.ts
@@ -250,6 +250,38 @@ describe("operator dashboard payload", () => {
     expect(JSON.stringify(payload)).not.toContain("missing_linked_issue"); // the healthy rule never appears
   });
 
+  it("surfaces the per-project rule breakdown (#8906), catching a rule broken on ONE repo even while a second repo's healthy usage pools it back above the blended floor", async () => {
+    const env = createTestEnv();
+    const seedClose = async (id: string, project: string, ruleCode: string, truth: "closed" | "merged"): Promise<void> => {
+      await env.DB.prepare(
+        `INSERT INTO review_audit (id, project, target_id, event_type, decision, summary, source, created_at) VALUES (?, ?, ?, 'gate_decision', 'close', ?, 'gittensory-native', ?)`,
+      )
+        .bind(`gd-${id}`, project, `${project}#${id}`, ruleCode, new Date().toISOString())
+        .run();
+      await env.DB.prepare(
+        `INSERT INTO review_audit (id, project, target_id, event_type, decision, source, created_at) VALUES (?, ?, ?, 'pr_outcome', ?, 'github', ?)`,
+      )
+        .bind(`po-${id}`, project, `${project}#${id}`, truth, new Date().toISOString())
+        .run();
+    };
+    // Broken on ONE repo: 12 closes (clears AUTOTUNE_MIN_DECIDED), every single one later merged -- 0% precision.
+    for (let i = 1; i <= 12; i++) await seedClose(`bad-${i}`, "metagraphed/broken-repo", "shared_rule", "merged");
+    // The SAME rule, healthy on a SECOND repo: 90 closes, all correctly closed -- pools the blended precision to
+    // 90/102 ≈ 0.882, above AUTOTUNE_CLOSE_PRECISION_FLOOR (0.85), so the blended tile does NOT flag this rule
+    // even though it is genuinely broken on metagraphed/broken-repo -- exactly the gap #8906 exists to close.
+    for (let i = 1; i <= 90; i++) await seedClose(`good-${i}`, "metagraphed/healthy-repo", "shared_rule", "closed");
+
+    const payload = await buildOperatorDashboardPayload(env);
+    expect(payload.metrics).toEqual(
+      expect.arrayContaining([expect.objectContaining({ label: "Rules below close-precision floor", value: "0", delta: "no rule below floor" })]),
+    );
+    expect(payload.metrics).toEqual(
+      expect.arrayContaining([
+        expect.objectContaining({ label: "Rules below close-precision floor (per-project)", value: "1", delta: "metagraphed/broken-repo:shared_rule" }),
+      ]),
+    );
+  });
+
   it("wires computeFindingAcceptance into the dashboard's acceptance card shape (#1967/#5213)", async () => {
     const env = createTestEnv();
     await env.DB.prepare(
diff --git a/test/unit/rule-gate-eval.test.ts b/test/unit/rule-gate-eval.test.ts
index 345f3c19b7..004ef6ed77 100644
--- a/test/unit/rule-gate-eval.test.ts
+++ b/test/unit/rule-gate-eval.test.ts
@@ -3,6 +3,7 @@ import {
   computeRuleGateEval,
   computeBlendedRuleGateEval,
   rulesBelowClosePrecisionFloor,
+  projectRulesBelowClosePrecisionFloor,
   type RuleGateEvalRow,
   type BlendedRuleGateEvalRow,
 } from "../../src/review/rule-gate-eval";
@@ -383,3 +384,68 @@ describe("rulesBelowClosePrecisionFloor (#7984, #7986's own read)", () => {
     expect(rulesBelowClosePrecisionFloor(justBelow)).toHaveLength(1);
   });
 });
+
+function projectRow(overrides: Partial<RuleGateEvalRow> = {}): RuleGateEvalRow {
+  return {
+    project: "acme/widgets",
+    ruleCode: "rule_a",
+    wouldMerge: 0,
+    mergeConfirmed: 0,
+    mergeFalse: 0,
+    wouldClose: 0,
+    closeConfirmed: 0,
+    closeFalse: 0,
+    decided: 0,
+    mergePrecision: null,
+    closePrecision: null,
+    weightedMergeConfirmed: 0,
+    weightedCloseConfirmed: 0,
+    weightedMergePrecision: null,
+    weightedClosePrecision: null,
+    ...overrides,
+  };
+}
+
+describe("projectRulesBelowClosePrecisionFloor — per-(project, ruleCode) counterpart (#8906)", () => {
+  it("flags a rule once its sample clears AUTOTUNE_MIN_DECIDED with a below-floor weighted precision, on the SPECIFIC project it fired on", () => {
+    const rows = [
+      projectRow({ project: "acme/widgets", ruleCode: "surface_lane_reject", wouldClose: 12, closeConfirmed: 0, weightedClosePrecision: 0 }),
+    ];
+    const flagged = projectRulesBelowClosePrecisionFloor(rows, AUTOTUNE_CLOSE_PRECISION_FLOOR, AUTOTUNE_MIN_DECIDED);
+    expect(flagged).toHaveLength(1);
+    expect(flagged[0]?.project).toBe("acme/widgets");
+    expect(flagged[0]?.ruleCode).toBe("surface_lane_reject");
+  });
+
+  it("does NOT flag a rule with insufficient sample even at 0% precision, mirroring the blended fold's same floor", () => {
+    const rows = [projectRow({ ruleCode: "rare_rule", wouldClose: 3, closeConfirmed: 0, weightedClosePrecision: 0 })];
+    expect(projectRulesBelowClosePrecisionFloor(rows, AUTOTUNE_CLOSE_PRECISION_FLOOR, AUTOTUNE_MIN_DECIDED)).toEqual([]);
+  });
+
+  it("does NOT flag a healthy rule with plenty of sample and good precision", () => {
+    const rows = [projectRow({ ruleCode: "healthy_rule", wouldClose: 40, closeConfirmed: 39, weightedClosePrecision: 39 / 40 })];
+    expect(projectRulesBelowClosePrecisionFloor(rows, AUTOTUNE_CLOSE_PRECISION_FLOOR, AUTOTUNE_MIN_DECIDED)).toEqual([]);
+  });
+
+  it("does NOT flag a rule whose weightedClosePrecision is null (no decided close verdict at all)", () => {
+    const rows = [projectRow({ ruleCode: "merge_only_rule", wouldClose: 0, weightedClosePrecision: null })];
+    expect(projectRulesBelowClosePrecisionFloor(rows, AUTOTUNE_CLOSE_PRECISION_FLOOR, AUTOTUNE_MIN_DECIDED)).toEqual([]);
+  });
+
+  it("defaults floor and minDecided to the SAME constants the blended fold uses", () => {
+    const atFloor = [projectRow({ ruleCode: "rule_a", wouldClose: 20, weightedClosePrecision: AUTOTUNE_CLOSE_PRECISION_FLOOR })];
+    expect(projectRulesBelowClosePrecisionFloor(atFloor)).toEqual([]);
+    const justBelow = [projectRow({ ruleCode: "rule_a", wouldClose: 20, weightedClosePrecision: AUTOTUNE_CLOSE_PRECISION_FLOOR - 0.01 })];
+    expect(projectRulesBelowClosePrecisionFloor(justBelow)).toHaveLength(1);
+  });
+
+  it("isolates a rule broken on ONE project even when a second project's healthy rows would pool it away in the blended view", () => {
+    const rows = [
+      projectRow({ project: "acme/broken-repo", ruleCode: "shared_rule", wouldClose: 15, closeConfirmed: 0, weightedClosePrecision: 0 }),
+      projectRow({ project: "acme/healthy-repo", ruleCode: "shared_rule", wouldClose: 30, closeConfirmed: 30, weightedClosePrecision: 1 }),
+    ];
+    const flagged = projectRulesBelowClosePrecisionFloor(rows, AUTOTUNE_CLOSE_PRECISION_FLOOR, AUTOTUNE_MIN_DECIDED);
+    expect(flagged).toHaveLength(1);
+    expect(flagged[0]?.project).toBe("acme/broken-repo");
+  });
+});