fix: normalize routed replies

CHANGELOG.md

@@ -7,7 +7,6 @@
 - Commands: accept /models as an alias for /model.
 - Debugging: add raw model stream logging flags and document gateway watch mode.
 - Agent: add claude-cli/opus-4.5 runner via Claude CLI with resume support (tools disabled).
-- Auth: respect cooldown tracking even with explicit `auth.order` (avoid repeatedly trying known-bad keys). — thanks @steipete
 - CLI: move `clawdbot message` to subcommands (`message send|poll|…`), fold Discord/Slack/Telegram/WhatsApp tools into `message`, and require `--provider` unless only one provider is configured.
 - CLI: improve `logs` output (pretty/plain/JSONL), add gateway unreachable hint, and document logging.
 - WhatsApp: route queued replies to the original sender instead of the bot's own number. (#534) — thanks @mcinteerj
@@ -15,7 +14,6 @@
 - Deps: bump Pi to 0.40.0 and drop pi-ai patch (upstream 429 fix). (#543) — thanks @mcinteerj
 - Security: per-agent mention patterns and group elevated directives now require explicit mention to avoid cross-agent toggles.
 - Config: support inline env vars in config (`env.*` / `env.vars`) and document env precedence.
-- Config: migrate routing/agent config into agents.list/agents.defaults and messages/tools/audio with default agent selection and per-agent identity config.
 - Agent: enable adaptive context pruning by default for tool-result trimming.
 - Doctor: check config/state permissions and offer to tighten them. — thanks @steipete
 - Doctor/Daemon: audit supervisor configs, add --repair/--force flows, surface service config audits in daemon status, and document user vs system services. — thanks @steipete
@@ -58,7 +56,7 @@
 - Onboarding: QuickStart jumps straight into provider selection with Telegram preselected when unset.
 - Onboarding: QuickStart auto-installs the Gateway daemon with Node (no runtime picker).
 - Onboarding: clarify WhatsApp owner number prompt and label pairing phone number.
-- Onboarding: add hosted MiniMax M2.1 API key flow + config. (#495) — thanks @tobiasbischoff
+- Auto-reply: normalize routed replies to drop NO_REPLY and apply response prefixes.
 - Daemon runtime: remove Bun from selection options.
 - CLI: restore hidden `gateway-daemon` alias for legacy launchd configs.
 - Onboarding/Configure: add OpenAI API key flow that stores in shared `~/.clawdbot/.env` for launchd; simplify Anthropic token prompt order.
@@ -111,8 +109,8 @@
   - New default: DM pairing (`dmPolicy="pairing"` / `discord.dm.policy="pairing"` / `slack.dm.policy="pairing"`).
   - To keep old “open to everyone” behavior: set `dmPolicy="open"` and include `"*"` in the relevant `allowFrom` (Discord/Slack: `discord.dm.allowFrom` / `slack.dm.allowFrom`).
   - Approve requests via `clawdbot pairing list --provider <provider>` + `clawdbot pairing approve --provider <provider> <code>`.
-- Sandbox: default `agents.defaults.sandbox.scope` to `"agent"` (one container/workspace per agent). Use `"session"` for per-session isolation; `"shared"` disables cross-session isolation.
+- Sandbox: default `agent.sandbox.scope` to `"agent"` (one container/workspace per agent). Use `"session"` for per-session isolation; `"shared"` disables cross-session isolation.
-- Timestamps in agent envelopes are now UTC (compact `YYYY-MM-DDTHH:mmZ`); removed `messages.timestampPrefix`. Add `agents.defaults.userTimezone` to tell the model the user’s local time (system prompt only).
+- Timestamps in agent envelopes are now UTC (compact `YYYY-MM-DDTHH:mmZ`); removed `messages.timestampPrefix`. Add `agent.userTimezone` to tell the model the user’s local time (system prompt only).
 - Model config schema changes (auth profiles + model lists); doctor auto-migrates and the gateway rewrites legacy configs on startup.
 - Commands: gate all slash commands to authorized senders; add `/compact` to manually compact session context.
 - Groups: `whatsapp.groups`, `telegram.groups`, and `imessage.groups` now act as allowlists when set. Add `"*"` to keep allow-all behavior.
@@ -138,7 +136,7 @@
 ## 2026.1.5
 
 ### Highlights
-- Models: add image-specific model config (`agents.defaults.imageModel` + fallbacks) and scan support.
+- Models: add image-specific model config (`agent.imageModel` + fallbacks) and scan support.
 - Agent tools: new `image` tool routed to the image model (when configured).
 - Config: default model shorthands (`opus`, `sonnet`, `gpt`, `gpt-mini`, `gemini`, `gemini-flash`).
 - Docs: document built-in model shorthands + precedence (user config wins).
@@ -163,7 +161,7 @@
 - Agent tools: OpenAI-compatible tool JSON Schemas (fix `browser`, normalize union schemas).
 - Onboarding: when running from source, auto-build missing Control UI assets (`bun run ui:build`).
 - Discord/Slack: route reaction + system notifications to the correct session (no main-session bleed).
-- Agent tools: honor `tools.allow` / `tools.deny` policy even when sandbox is off.
+- Agent tools: honor `agent.tools` allow/deny policy even when sandbox is off.
 - Discord: avoid duplicate replies when OpenAI emits repeated `message_end` events.
 - Commands: unify /status (inline) and command auth across providers; group bypass for authorized control commands; remove Discord /clawd slash handler.
 - CLI: run `clawdbot agent` via the Gateway by default; use `--local` to force embedded mode.

src/auto-reply/reply/normalize-reply.ts

@@ -0,0 +1,49 @@
+import { stripHeartbeatToken } from "../heartbeat.js";
+import { HEARTBEAT_TOKEN, SILENT_REPLY_TOKEN } from "../tokens.js";
+import type { ReplyPayload } from "../types.js";
+
+export type NormalizeReplyOptions = {
+  responsePrefix?: string;
+  onHeartbeatStrip?: () => void;
+  stripHeartbeat?: boolean;
+  silentToken?: string;
+};
+
+export function normalizeReplyPayload(
+  payload: ReplyPayload,
+  opts: NormalizeReplyOptions = {},
+): ReplyPayload | null {
+  const hasMedia = Boolean(
+    payload.mediaUrl || (payload.mediaUrls?.length ?? 0) > 0,
+  );
+  const trimmed = payload.text?.trim() ?? "";
+  if (!trimmed && !hasMedia) return null;
+
+  const silentToken = opts.silentToken ?? SILENT_REPLY_TOKEN;
+  if (trimmed === silentToken && !hasMedia) return null;
+
+  let text = payload.text ?? undefined;
+  if (text && !trimmed) {
+    // Keep empty text when media exists so media-only replies still send.
+    text = "";
+  }
+
+  const shouldStripHeartbeat = opts.stripHeartbeat ?? true;
+  if (shouldStripHeartbeat && text?.includes(HEARTBEAT_TOKEN)) {
+    const stripped = stripHeartbeatToken(text, { mode: "message" });
+    if (stripped.didStrip) opts.onHeartbeatStrip?.();
+    if (stripped.shouldSkip && !hasMedia) return null;
+    text = stripped.text;
+  }
+
+  if (
+    opts.responsePrefix &&
+    text &&
+    text.trim() !== HEARTBEAT_TOKEN &&
+    !text.startsWith(opts.responsePrefix)
+  ) {
+    text = `${opts.responsePrefix} ${text}`;
+  }
+
+  return { ...payload, text };
+}

src/auto-reply/reply/reply-dispatcher.ts

@@ -1,5 +1,4 @@
-import { stripHeartbeatToken } from "../heartbeat.js";
+import { normalizeReplyPayload } from "./normalize-reply.js";
-import { HEARTBEAT_TOKEN, SILENT_REPLY_TOKEN } from "../tokens.js";
 import type { GetReplyOptions, ReplyPayload } from "../types.js";
 import type { TypingController } from "./typing.js";
 
@@ -45,41 +44,14 @@ export type ReplyDispatcher = {
   getQueuedCounts: () => Record<ReplyDispatchKind, number>;
 };
 
-function normalizeReplyPayload(
+function normalizeReplyPayloadInternal(
   payload: ReplyPayload,
   opts: Pick<ReplyDispatcherOptions, "responsePrefix" | "onHeartbeatStrip">,
 ): ReplyPayload | null {
-  const hasMedia = Boolean(
+  return normalizeReplyPayload(payload, {
-    payload.mediaUrl || (payload.mediaUrls?.length ?? 0) > 0,
+    responsePrefix: opts.responsePrefix,
-  );
+    onHeartbeatStrip: opts.onHeartbeatStrip,
-  const trimmed = payload.text?.trim() ?? "";
+  });
-  if (!trimmed && !hasMedia) return null;
-
-  // Avoid sending the explicit silent token when no media is attached.
-  if (trimmed === SILENT_REPLY_TOKEN && !hasMedia) return null;
-
-  let text = payload.text ?? undefined;
-  if (text && !trimmed) {
-    // Keep empty text when media exists so media-only replies still send.
-    text = "";
-  }
-  if (text?.includes(HEARTBEAT_TOKEN)) {
-    const stripped = stripHeartbeatToken(text, { mode: "message" });
-    if (stripped.didStrip) opts.onHeartbeatStrip?.();
-    if (stripped.shouldSkip && !hasMedia) return null;
-    text = stripped.text;
-  }
-
-  if (
-    opts.responsePrefix &&
-    text &&
-    text.trim() !== HEARTBEAT_TOKEN &&
-    !text.startsWith(opts.responsePrefix)
-  ) {
-    text = `${opts.responsePrefix} ${text}`;
-  }
-
-  return { ...payload, text };
 }
 
 export function createReplyDispatcher(
@@ -96,7 +68,7 @@ export function createReplyDispatcher(
   };
 
   const enqueue = (kind: ReplyDispatchKind, payload: ReplyPayload) => {
-    const normalized = normalizeReplyPayload(payload, options);
+    const normalized = normalizeReplyPayloadInternal(payload, options);
     if (!normalized) return false;
     queuedCounts[kind] += 1;
     pending += 1;

src/auto-reply/reply/route-reply.test.ts

@@ -1,6 +1,7 @@
 import { describe, expect, it, vi } from "vitest";
 
 import type { ClawdbotConfig } from "../../config/config.js";
+import { SILENT_REPLY_TOKEN } from "../tokens.js";
 
 const mocks = vi.hoisted(() => ({
   sendMessageDiscord: vi.fn(async () => ({ messageId: "m1", channelId: "c1" })),
@@ -68,6 +69,36 @@ describe("routeReply", () => {
     expect(mocks.sendMessageSlack).not.toHaveBeenCalled();
   });
 
+  it("drops silent token payloads", async () => {
+    mocks.sendMessageSlack.mockClear();
+    const res = await routeReply({
+      payload: { text: SILENT_REPLY_TOKEN },
+      channel: "slack",
+      to: "channel:C123",
+      cfg: {} as never,
+    });
+    expect(res.ok).toBe(true);
+    expect(mocks.sendMessageSlack).not.toHaveBeenCalled();
+  });
+
+  it("applies responsePrefix when routing", async () => {
+    mocks.sendMessageSlack.mockClear();
+    const cfg = {
+      messages: { responsePrefix: "[clawdbot]" },
+    } as unknown as ClawdbotConfig;
+    await routeReply({
+      payload: { text: "hi" },
+      channel: "slack",
+      to: "channel:C123",
+      cfg,
+    });
+    expect(mocks.sendMessageSlack).toHaveBeenCalledWith(
+      "channel:C123",
+      "[clawdbot] hi",
+      expect.any(Object),
+    );
+  });
+
   it("passes thread id to Telegram sends", async () => {
     mocks.sendMessageTelegram.mockClear();
     await routeReply({

src/auto-reply/reply/route-reply.ts

@@ -17,6 +17,7 @@ import { sendMessageTelegram } from "../../telegram/send.js";
 import { sendMessageWhatsApp } from "../../web/outbound.js";
 import type { OriginatingChannelType } from "../templating.js";
 import type { ReplyPayload } from "../types.js";
+import { normalizeReplyPayload } from "./normalize-reply.js";
 
 export type RouteReplyParams = {
   /** The reply payload to send. */
@@ -59,13 +60,18 @@ export async function routeReply(
     params;
 
   // Debug: `pnpm test src/auto-reply/reply/route-reply.test.ts`
-  const text = payload.text ?? "";
+  const normalized = normalizeReplyPayload(payload, {
-  const mediaUrls = (payload.mediaUrls?.filter(Boolean) ?? []).length
+    responsePrefix: cfg.messages?.responsePrefix,
-    ? (payload.mediaUrls?.filter(Boolean) as string[])
+  });
-    : payload.mediaUrl
+  if (!normalized) return { ok: true };
-      ? [payload.mediaUrl]
+
+  const text = normalized.text ?? "";
+  const mediaUrls = (normalized.mediaUrls?.filter(Boolean) ?? []).length
+    ? (normalized.mediaUrls?.filter(Boolean) as string[])
+    : normalized.mediaUrl
+      ? [normalized.mediaUrl]
       : [];
-  const replyToId = payload.replyToId;
+  const replyToId = normalized.replyToId;
 
   // Skip empty replies.
   if (!text.trim() && mediaUrls.length === 0) {