fix(webchat): keep trustedLocalMedia internal to reply payloads

Restore Omit on public plugin-sdk ReplyPayload; set trustedLocalMedia via
runtime assertion in speech-core and explicitly on dispatch TTS-only finals.

Co-authored-by: Cursor <cursoragent@cursor.com>

extensions/speech-core/src/tts.test.ts

@@ -210,7 +210,7 @@ async function expectTtsPayloadResult(params: {
     expect(result.audioAsVoice).toBe(params.audioAsVoice);
     expect(result.mediaUrl).toMatch(new RegExp(`voice-\\d+\\.${params.mediaExtension ?? "ogg"}$`));
     expect(result.spokenText).toBe(params.text);
-    expect(result.trustedLocalMedia).toBe(true);
+    expect((result as { trustedLocalMedia?: boolean }).trustedLocalMedia).toBe(true);
 
     mediaDir = result.mediaUrl ? path.dirname(result.mediaUrl) : undefined;
   } finally {
@@ -453,7 +453,7 @@ describe("speech-core native voice-note routing", () => {
     });
 
     expect(synthesizeMock).not.toHaveBeenCalled();
-    expect(result.trustedLocalMedia).toBeUndefined();
+    expect((result as { trustedLocalMedia?: boolean }).trustedLocalMedia).toBeUndefined();
     expect(result.text).toBe("WebChat block stream chunks defer TTS to the final tail.");
   });
 
@@ -468,7 +468,7 @@ describe("speech-core native voice-note routing", () => {
     });
 
     expect(synthesizeMock).not.toHaveBeenCalled();
-    expect(result.trustedLocalMedia).toBeUndefined();
+    expect((result as { trustedLocalMedia?: boolean }).trustedLocalMedia).toBeUndefined();
     expect(result.text).toBe("Intermediate tool output should not be spoken.");
   });
 

extensions/speech-core/src/tts.ts

@@ -1850,7 +1850,7 @@ export async function maybeApplyTtsToPayload(params: {
       audioAsVoice: result.audioAsVoice || params.payload.audioAsVoice,
       spokenText: textForAudio,
       trustedLocalMedia: true,
-    };
+    } as ReplyPayload;
   }
 
   lastTtsAttempt = {

scripts/repro/webchat-auto-tts-live-proof.mjs

@@ -89,16 +89,14 @@ async function main() {
     throw new Error("expected final-mode tail TTS to write a local media file");
   }
 
+  // Same shape as dispatch-from-config accumulated block TTS-only final payload.
   const ttsOnlyPayload = {
     mediaUrl: tailResult.mediaUrl,
     audioAsVoice: tailResult.audioAsVoice,
     spokenText: accumulatedBlockText,
-    trustedLocalMedia: tailResult.trustedLocalMedia,
+    trustedLocalMedia: true,
   };
-  console.log(
-    "dispatch ttsOnlyPayload.trustedLocalMedia =",
-    ttsOnlyPayload.trustedLocalMedia ?? false,
-  );
+  console.log("dispatch ttsOnlyPayload.trustedLocalMedia =", ttsOnlyPayload.trustedLocalMedia);
 
   const localRoots = [path.dirname(mediaPath)];
   const trustedBlocks = await buildWebchatAudioContentBlocksFromReplyPayloads([ttsOnlyPayload], {

src/auto-reply/reply/dispatch-from-config.ts

@@ -1700,7 +1700,7 @@ export async function dispatchReplyFromConfig(
               mediaUrl: ttsSyntheticReply.mediaUrl,
               audioAsVoice: ttsSyntheticReply.audioAsVoice,
               spokenText: accumulatedBlockTtsText,
-              trustedLocalMedia: ttsSyntheticReply.trustedLocalMedia,
+              trustedLocalMedia: true,
             };
             const normalizedTtsOnlyPayload = await normalizeReplyMediaPayload(ttsOnlyPayload);
             const result = await routeReplyToOriginating(normalizedTtsOnlyPayload);

src/plugin-sdk/reply-payload.ts

@@ -6,7 +6,7 @@ import { normalizeLowercaseStringOrEmpty, readStringValue } from "../shared/stri
 
 export type { MediaPayload, MediaPayloadInput } from "../channels/plugins/media-payload.js";
 export { buildMediaPayload } from "../channels/plugins/media-payload.js";
-export type ReplyPayload = InternalReplyPayload;
+export type ReplyPayload = Omit<InternalReplyPayload, "trustedLocalMedia">;
 
 export type OutboundReplyPayload = {
   text?: string;