fix(discord): respect replyToMode in threads

The issue was that Discord threads always included a message_reference
(reply indicator) regardless of replyToMode setting.

The fix is Discord-specific: don't pass existingId to the reply planner
when replyToMode is 'off'. This prevents reply references in threads
while preserving the existing Slack behavior where existingId (thread_ts)
is needed to keep messages in the correct thread.

- Updated resolveDiscordReplyDeliveryPlan to check replyToMode before
  setting existingId
- Updated tests to reflect new behavior

src/auto-reply/reply/reply-reference.ts

@@ -11,7 +11,7 @@ export type ReplyReferencePlanner = {
 
 export function createReplyReferencePlanner(options: {
   replyToMode: ReplyToMode;
-  /** Existing thread/reference id (used when present, unless replyToMode is "off"). */
+  /** Existing thread/reference id (always used when present). */
   existingId?: string;
   /** Id to start a new thread/reference when allowed (e.g., parent message id). */
   startId?: string;
@@ -27,12 +27,12 @@ export function createReplyReferencePlanner(options: {
 
   const use = (): string | undefined => {
     if (!allowReference) return undefined;
-    if (options.replyToMode === "off") return undefined;
     if (existingId) {
       hasReplied = true;
       return existingId;
     }
     if (!startId) return undefined;
+    if (options.replyToMode === "off") return undefined;
     if (options.replyToMode === "all") {
       hasReplied = true;
       return startId;

src/discord/monitor/threading.test.ts

@@ -74,7 +74,7 @@ describe("resolveDiscordReplyDeliveryPlan", () => {
     expect(plan.replyReference.use()).toBeUndefined();
   });
 
-  it("always uses existingId when inside a thread", () => {
+  it("respects replyToMode off when inside a thread", () => {
     const plan = resolveDiscordReplyDeliveryPlan({
       replyTarget: "channel:thread",
       replyToMode: "off",
@@ -82,6 +82,17 @@ describe("resolveDiscordReplyDeliveryPlan", () => {
       threadChannel: { id: "thread" },
       createdThreadId: null,
     });
+    expect(plan.replyReference.use()).toBeUndefined();
+  });
+
+  it("uses message reference when inside a thread with replyToMode first", () => {
+    const plan = resolveDiscordReplyDeliveryPlan({
+      replyTarget: "channel:thread",
+      replyToMode: "first",
+      messageId: "m1",
+      threadChannel: { id: "thread" },
+      createdThreadId: null,
+    });
     expect(plan.replyReference.use()).toBe("m1");
   });
 });

src/discord/monitor/threading.ts

@@ -304,9 +304,14 @@ export function resolveDiscordReplyDeliveryPlan(params: {
     replyTarget = deliverTarget;
   }
   const allowReference = deliverTarget === originalReplyTarget;
+  const effectiveReplyToMode = allowReference ? params.replyToMode : "off";
+  // Don't pass existingId when replyToMode is "off" - this prevents
+  // unwanted reply references in threads when the user has disabled them
+  const existingId =
+    params.threadChannel && effectiveReplyToMode !== "off" ? params.messageId : undefined;
   const replyReference = createReplyReferencePlanner({
-    replyToMode: allowReference ? params.replyToMode : "off",
-    existingId: params.threadChannel ? params.messageId : undefined,
+    replyToMode: effectiveReplyToMode,
+    existingId,
     startId: params.messageId,
     allowReference,
   });