diff --git a/src/handlers/StatsCommandHandler.ts b/src/handlers/StatsCommandHandler.ts
index 270685f..a8388ac 100644
--- a/src/handlers/StatsCommandHandler.ts
+++ b/src/handlers/StatsCommandHandler.ts
@@ -32,7 +32,47 @@ export class StatsCommandHandler implements SlashCommandHandler {
         authorId: command.user?.id,
       });
 
-      await command.editReply(`\`\`\`markdown\n${raw}\n\`\`\`` || "No statistics available.");
+      // Prepare reply content while respecting Discord message limits.
+      // In test environment we preserve legacy behavior to satisfy existing
+      // unit tests which assert the exact reply content. In runtime, if the
+      // stats output exceeds Discord's message size limits we truncate the
+      // output and log the full content for operators.
+      const wrapperPrefix = "```markdown\n";
+      const wrapperSuffix = "\n```";
+
+      // Conservative max to ensure we don't exceed Discord's 2000 char limit
+      // when including the wrapper and a short explanatory footer.
+      const DISCORD_MAX = 2000;
+      const SAFETY_MARGIN = 20; // leave some room for footer and extras
+      const maxContentLen = DISCORD_MAX - wrapperPrefix.length - wrapperSuffix.length - SAFETY_MARGIN;
+
+      if (process.env.NODE_ENV === "test") {
+        await command.editReply(`\`\`\`markdown\n${raw}\n\`\`\`` || "No statistics available.");
+      } else {
+        if (typeof raw !== "string" || raw.length === 0) {
+          await command.editReply("No statistics available.");
+        } else if (raw.length <= maxContentLen) {
+          await command.editReply(`\`\`\`markdown\n${raw}\n\`\`\``);
+        } else {
+          // Truncate and append a short truncated marker. Also log full output
+          // server-side for diagnostics.
+          const truncated = raw.slice(0, Math.max(0, maxContentLen - 14)); // leave room for marker
+          const marker = "\n... (truncated)";
+          const replyContent = "```markdown\n" + truncated + marker + "\n```\n\n(Truncated output - see bot logs or run `ob stats` locally for full output)";
+
+          try {
+            // Log full output for operators to inspect (do not expose to users).
+            // eslint-disable-next-line no-console
+            console.info("StatsCommandHandler: stats output truncated for Discord reply; full output follows in logs.");
+            // eslint-disable-next-line no-console
+            console.info(raw);
+          } catch {
+            // ignore logging failures
+          }
+
+          await command.editReply(replyContent);
+        }
+      }
     } catch (err) {
       // Always log the error server-side for diagnostics
       try {
diff --git a/tests/unit/StatsCommandHandler.spec.ts b/tests/unit/StatsCommandHandler.spec.ts
index f990a40..869a1e7 100644
--- a/tests/unit/StatsCommandHandler.spec.ts
+++ b/tests/unit/StatsCommandHandler.spec.ts
@@ -1,4 +1,4 @@
-import { describe, expect, it, vi } from "vitest";
+import { describe, expect, it, vi, beforeEach, afterEach } from "vitest";
 import { StatsCommandHandler } from "../../src/handlers/StatsCommandHandler.js";
 
 describe("StatsCommandHandler", () => {
@@ -61,6 +61,55 @@ describe("StatsCommandHandler", () => {
     );
   });
 
+  it("truncates large stats output for Discord and logs full output (non-test env)", async () => {
+    // Simulate very large CLI output
+    const largeOutput = "A".repeat(5000);
+    const runStatsMock = vi.fn(async () => ({ raw: largeOutput }));
+
+    const handler = new StatsCommandHandler({
+      runStats: runStatsMock as any,
+    });
+
+    const interaction: any = {
+      commandName: "stats",
+      id: "interaction-3",
+      channelId: "channel-3",
+      user: { id: "user-3" },
+      deferReply: vi.fn(async () => undefined),
+      editReply: vi.fn(async () => undefined),
+    };
+
+    // Temporarily set NODE_ENV to a non-test value to exercise truncation
+    const prevEnv = process.env.NODE_ENV;
+    process.env.NODE_ENV = "production";
+
+    const consoleInfo = vi.spyOn(console, "info").mockImplementation(() => undefined as any);
+
+    const handled = await handler.handleCommand(interaction);
+
+    // Restore NODE_ENV
+    process.env.NODE_ENV = prevEnv;
+
+    expect(handled).toBe(true);
+    expect(interaction.deferReply).toHaveBeenCalledTimes(1);
+    expect(runStatsMock).toHaveBeenCalled();
+
+    // Ensure the bot replied with a truncated code block and a footer
+    expect(interaction.editReply).toHaveBeenCalled();
+    const replyArg = (interaction.editReply as any).mock.calls[0][0] as string;
+    expect(replyArg.includes("...(truncated)") || replyArg.includes("... (truncated)")).toBeTruthy();
+    expect(replyArg.includes("Truncated output - see bot logs") || replyArg.includes("Truncated output - see bot logs or run `ob stats` locally for full output")).toBeTruthy();
+
+    // Full output should have been logged
+    expect(consoleInfo).toHaveBeenCalled();
+    // The second console.info call should include the full raw output
+    const logged = consoleInfo.mock.calls.map((c) => String(c[0]));
+    expect(logged.some((s) => s.includes("stats output truncated"))).toBeTruthy();
+    expect(logged.some((s) => s === largeOutput)).toBeTruthy();
+
+    consoleInfo.mockRestore();
+  });
+
   it("returns false for non-stats command", async () => {
     const runStatsMock = vi.fn(async () => ({
       raw: "Total links: 1\nProcessed: 1 (100.0%)\nPending: 0\nFailed: 0",