Strip NO_REPLY token from mixed-content messages #30955
Description
Problem
When an agent sends a message like 😄 NO_REPLY, the isSilentReplyText check fails (it requires exact match with ^\s*NO_REPLY\s*$), so the entire message including the NO_REPLY token leaks through to the user as visible text.
Expected Behavior
NO_REPLYalone → suppress (current, works ✅)😄 NO_REPLY→ send😄(strip the token, deliver the content)😄→ send😄(no change needed)
Proposed Fix
In src/auto-reply/reply/normalize-reply.ts, after the isSilentReplyText exact-match check (~line 40), add a strip-and-keep fallback:
import { escapeRegExp } from "../../utils.js";
// Strip NO_REPLY token from mixed-content messages (e.g. "😄 NO_REPLY")
if (text && !isSilentReplyText(text, silentToken) && text.includes(silentToken)) {
text = text.replace(new RegExp(`\\s*${escapeRegExp(silentToken)}\\s*`, "g"), "").trim();
if (!text && !hasMedia && !hasChannelData) {
opts.onSkip?.("silent");
return null;
}
}Test cases for tokens.test.ts:
it("strips NO_REPLY from mixed emoji message", () => {
// This would be tested via normalizeReplyPayload, not isSilentReplyText
expect(isSilentReplyText("😄 NO_REPLY")).toBe(false); // correctly not silent
});Test cases for normalize-reply (new or existing test file):
it("strips NO_REPLY from mixed content and delivers remaining text", () => {
const result = normalizeReplyPayload({ text: "😄 NO_REPLY" });
expect(result).not.toBeNull();
expect(result!.text).toBe("😄");
});
it("strips NO_REPLY when it appears before content", () => {
const result = normalizeReplyPayload({ text: "NO_REPLY 👍" });
expect(result).not.toBeNull();
expect(result!.text).toBe("👍");
});
it("still suppresses pure NO_REPLY", () => {
const result = normalizeReplyPayload({ text: "NO_REPLY" });
expect(result).toBeNull();
});Context
Discovered in a Quiubo group chat when an agent sent 😄 NO_REPLY — the raw text including the token was delivered to all participants.