diff --git a/extensions/openai/image-generation-provider.test.ts b/extensions/openai/image-generation-provider.test.ts
new file mode 100644
index 00000000000..4321ef35098
--- /dev/null
+++ b/extensions/openai/image-generation-provider.test.ts
@@ -0,0 +1,83 @@
+import { afterEach, describe, expect, it, vi } from "vitest";
+import { buildOpenAIImageGenerationProvider } from "./image-generation-provider.js";
+
+const {
+  resolveApiKeyForProviderMock,
+  postJsonRequestMock,
+  postTranscriptionRequestMock,
+  assertOkOrThrowHttpErrorMock,
+  resolveProviderHttpRequestConfigMock,
+} = vi.hoisted(() => ({
+  resolveApiKeyForProviderMock: vi.fn(async () => ({ apiKey: "openai-key" })),
+  postJsonRequestMock: vi.fn(),
+  postTranscriptionRequestMock: vi.fn(),
+  assertOkOrThrowHttpErrorMock: vi.fn(async () => {}),
+  resolveProviderHttpRequestConfigMock: vi.fn((params) => ({
+    baseUrl: params.baseUrl ?? params.defaultBaseUrl,
+    allowPrivateNetwork: Boolean(params.allowPrivateNetwork ?? params.baseUrl?.trim()),
+    headers: new Headers(params.defaultHeaders),
+    dispatcherPolicy: undefined,
+  })),
+}));
+
+vi.mock("openclaw/plugin-sdk/provider-auth-runtime", () => ({
+  resolveApiKeyForProvider: resolveApiKeyForProviderMock,
+}));
+
+vi.mock("openclaw/plugin-sdk/provider-http", () => ({
+  assertOkOrThrowHttpError: assertOkOrThrowHttpErrorMock,
+  postJsonRequest: postJsonRequestMock,
+  postTranscriptionRequest: postTranscriptionRequestMock,
+  resolveProviderHttpRequestConfig: resolveProviderHttpRequestConfigMock,
+}));
+
+describe("openai image generation provider", () => {
+  afterEach(() => {
+    resolveApiKeyForProviderMock.mockClear();
+    postJsonRequestMock.mockReset();
+    postTranscriptionRequestMock.mockReset();
+    assertOkOrThrowHttpErrorMock.mockClear();
+    resolveProviderHttpRequestConfigMock.mockClear();
+  });
+
+  it("allows explicit local baseUrl overrides for image requests", async () => {
+    postJsonRequestMock.mockResolvedValue({
+      response: {
+        json: async () => ({
+          data: [{ b64_json: Buffer.from("png-bytes").toString("base64") }],
+        }),
+      },
+      release: vi.fn(async () => {}),
+    });
+
+    const provider = buildOpenAIImageGenerationProvider();
+    const result = await provider.generateImage({
+      provider: "openai",
+      model: "gpt-image-1",
+      prompt: "Draw a QA lighthouse",
+      cfg: {
+        models: {
+          providers: {
+            openai: {
+              baseUrl: "http://127.0.0.1:44080/v1",
+              models: [],
+            },
+          },
+        },
+      },
+    });
+
+    expect(resolveProviderHttpRequestConfigMock).toHaveBeenCalledWith(
+      expect.objectContaining({
+        baseUrl: "http://127.0.0.1:44080/v1",
+      }),
+    );
+    expect(postJsonRequestMock).toHaveBeenCalledWith(
+      expect.objectContaining({
+        url: "http://127.0.0.1:44080/v1/images/generations",
+        allowPrivateNetwork: true,
+      }),
+    );
+    expect(result.images).toHaveLength(1);
+  });
+});
diff --git a/extensions/openai/image-generation-provider.ts b/extensions/openai/image-generation-provider.ts
index 2f8a4035fe6..8a1a1a5e429 100644
--- a/extensions/openai/image-generation-provider.ts
+++ b/extensions/openai/image-generation-provider.ts
@@ -90,7 +90,6 @@ export function buildOpenAIImageGenerationProvider(): ImageGenerationProvider {
         resolveProviderHttpRequestConfig({
           baseUrl: resolveOpenAIBaseUrl(req.cfg),
           defaultBaseUrl: DEFAULT_OPENAI_IMAGE_BASE_URL,
-          allowPrivateNetwork: false,
           defaultHeaders: {
             Authorization: `Bearer ${auth.apiKey}`,
           },
diff --git a/extensions/qa-lab/src/mock-openai-server.test.ts b/extensions/qa-lab/src/mock-openai-server.test.ts
index f4447d19511..572b22db21e 100644
--- a/extensions/qa-lab/src/mock-openai-server.test.ts
+++ b/extensions/qa-lab/src/mock-openai-server.test.ts
@@ -251,6 +251,17 @@ describe("qa mock openai server", () => {
     expect(await image.json()).toMatchObject({
       data: [{ b64_json: expect.any(String) }],
     });
+
+    const imageRequests = await fetch(`${server.baseUrl}/debug/image-generations`);
+    expect(imageRequests.status).toBe(200);
+    expect(await imageRequests.json()).toMatchObject([
+      {
+        model: "gpt-image-1",
+        prompt: "Draw a QA lighthouse",
+        n: 1,
+        size: "1024x1024",
+      },
+    ]);
   });
 
   it("returns exact markers for visible and hot-installed skills", async () => {
diff --git a/extensions/qa-lab/src/mock-openai-server.ts b/extensions/qa-lab/src/mock-openai-server.ts
index b7f7557cddc..b555bcbde91 100644
--- a/extensions/qa-lab/src/mock-openai-server.ts
+++ b/extensions/qa-lab/src/mock-openai-server.ts
@@ -279,6 +279,11 @@ function extractOrbitCode(text: string) {
   return /\b(?:ORBIT-9|orbit-9)\b/.exec(text)?.[0]?.toUpperCase() ?? null;
 }
 
+function extractExactReplyDirective(text: string) {
+  const match = /reply with exactly:\s*([^\n]+)/i.exec(text);
+  return match?.[1]?.trim() || null;
+}
+
 function buildAssistantText(input: ResponsesInputItem[], body: Record<string, unknown>) {
   const prompt = extractLastUserText(input);
   const toolOutput = extractToolOutput(input);
@@ -295,6 +300,7 @@ function buildAssistantText(input: ResponsesInputItem[], body: Record<string, un
         : toolOutput;
   const orbitCode = extractOrbitCode(memorySnippet);
   const mediaPath = /MEDIA:([^\n]+)/.exec(toolOutput)?.[1]?.trim();
+  const exactReplyDirective = extractExactReplyDirective(allInputText);
 
   if (/what was the qa canary code/i.test(prompt) && rememberedFact) {
     return `Protocol note: the QA canary code was ${rememberedFact}.`;
@@ -305,6 +311,9 @@ function buildAssistantText(input: ResponsesInputItem[], body: Record<string, un
   if (/memory unavailable check/i.test(prompt)) {
     return "Protocol note: I checked the available runtime context but could not confirm the hidden memory-only fact, so I will not guess.";
   }
+  if (/\bmarker\b/i.test(prompt) && exactReplyDirective) {
+    return exactReplyDirective;
+  }
   if (/visible skill marker/i.test(prompt)) {
     return "VISIBLE-SKILL-OK";
   }
@@ -491,6 +500,7 @@ export async function startQaMockOpenAiServer(params?: { host?: string; port?: n
   const host = params?.host ?? "127.0.0.1";
   let lastRequest: MockOpenAiRequestSnapshot | null = null;
   const requests: MockOpenAiRequestSnapshot[] = [];
+  const imageGenerationRequests: Array<Record<string, unknown>> = [];
   const server = createServer(async (req, res) => {
     const url = new URL(req.url ?? "/", "http://127.0.0.1");
     if (req.method === "GET" && (url.pathname === "/healthz" || url.pathname === "/readyz")) {
@@ -515,7 +525,17 @@ export async function startQaMockOpenAiServer(params?: { host?: string; port?: n
       writeJson(res, 200, requests);
       return;
     }
+    if (req.method === "GET" && url.pathname === "/debug/image-generations") {
+      writeJson(res, 200, imageGenerationRequests);
+      return;
+    }
     if (req.method === "POST" && url.pathname === "/v1/images/generations") {
+      const raw = await readBody(req);
+      const body = raw ? (JSON.parse(raw) as Record<string, unknown>) : {};
+      imageGenerationRequests.push(body);
+      if (imageGenerationRequests.length > 20) {
+        imageGenerationRequests.splice(0, imageGenerationRequests.length - 20);
+      }
       writeJson(res, 200, {
         data: [
           {