fix: no retry on quota limit errors (#1072)

src/tests/cli/errorFormatter.test.ts

@@ -182,6 +182,60 @@ describe("formatErrorDetails", () => {
     expect(message).toContain("/model");
   });
 
+  test("uses premium-specific guidance for premium-usage-exceeded", () => {
+    const error = new APIError(
+      429,
+      {
+        error: "Rate limited",
+        reasons: ["premium-usage-exceeded"],
+      },
+      undefined,
+      new Headers(),
+    );
+
+    const message = formatErrorDetails(error);
+
+    expect(message).toContain("Premium model usage limit");
+    expect(message).toContain("Standard or Basic hosted models");
+    expect(message).toContain("/model");
+    expect(message).not.toContain("hosted model usage limit");
+  });
+
+  test("uses standard-specific guidance for standard-usage-exceeded", () => {
+    const error = new APIError(
+      429,
+      {
+        error: "Rate limited",
+        reasons: ["standard-usage-exceeded"],
+      },
+      undefined,
+      new Headers(),
+    );
+
+    const message = formatErrorDetails(error);
+
+    expect(message).toContain("Standard model usage limit");
+    expect(message).toContain("Basic hosted models");
+    expect(message).toContain("/model");
+  });
+
+  test("uses basic-specific guidance for basic-usage-exceeded", () => {
+    const error = new APIError(
+      429,
+      {
+        error: "Rate limited",
+        reasons: ["basic-usage-exceeded"],
+      },
+      undefined,
+      new Headers(),
+    );
+
+    const message = formatErrorDetails(error);
+
+    expect(message).toContain("Basic model usage limit");
+    expect(message).toContain("/model");
+  });
+
   test("formats Z.ai error from APIError with embedded error code", () => {
     const error = new APIError(
       429,

src/tests/turn-recovery-policy.test.ts

@@ -195,6 +195,18 @@ describe("provider detail retry helpers", () => {
         "OpenAI API error: invalid_request_error",
       ),
     ).toBe(false);
+    expect(
+      shouldRetryRunMetadataError(
+        "llm_error",
+        '429 {"error":"Rate limited","reasons":["exceeded-quota"]}',
+      ),
+    ).toBe(false);
+    expect(
+      shouldRetryRunMetadataError(
+        "llm_error",
+        "You've reached your hosted model usage limit.",
+      ),
+    ).toBe(false);
   });
 
   test("pre-stream transient classifier handles status and detail", () => {
@@ -218,6 +230,13 @@ describe("provider detail retry helpers", () => {
           '429 {"error":"Rate limited","reasons":["agents-limit-exceeded"]}',
       }),
     ).toBe(false);
+    expect(
+      shouldRetryPreStreamTransientError({
+        status: 429,
+        detail:
+          '429 {"error":"Rate limited","reasons":["premium-usage-exceeded"]}',
+      }),
+    ).toBe(false);
     expect(
       shouldRetryPreStreamTransientError({
         status: 401,