Nohello

#nohello

Background: nohello.org and related sites document a common chat anti-pattern — opening with a bare "Hi" or "Hello" and waiting for a reply before stating the actual question. That forces synchronous back-and-forth and wastes the other person's time, especially across time zones. This skill operationalizes the fix: wait briefly for the real message, and if it doesn't show up, nudge the sender to just ask.

This skill has two modes of use:

1. Inline, as Claude — when you (Claude) are the one about to reply to a message that is only a greeting, follow the detection and response guidance below directly.
2. As an automated bot — for a persistent chat integration (Slack, Teams, Zoom, LinkedIn, or any other chat engine) that needs a real wait-then-nudge timer running outside of a single model turn, wire up the @nohello/core engine from this repository (https://github.com/ji-in-loop/nohello) instead of trying to reproduce the timer in-prompt. This SKILL.md and that package share the same detection rules and config shape, so behavior stays consistent whichever mode is in play.

1. Detect a greeting-only message

A message counts as greeting-only when, after removing:

• a greeting opener (hi, hello, hey, hiya, howdy, greetings, good morning/afternoon/evening, yo, sup, what's up, ...), and
• small-talk filler (how are you, how's it going, hope you're well, long time no see, you there?, got a sec?, quick question, ...), and
• a trailing name or address term (Hello Bala, Hi team),

there are no more than max_leftover_words words left (default 2), and the original message is no longer than max_greeting_words words (default 8). A message with a real ask attached — e.g. "Hi, can you review PR #123 today?" — is not greeting-only and should get a normal reply, not a nudge.

The exact algorithm is implemented in packages/core/src/detector.ts (detectGreetingOnly) — treat that as the source of truth if this description and the code ever disagree.

2. Wait for the real question

Default wait: 90 seconds (wait_seconds, configurable). If the same sender, in the same conversation, sends a substantive follow-up before the wait elapses, do nothing further — they answered their own #nohello. If another greeting-only message arrives, restart the wait.

In-prompt (mode 1), you generally can't literally sleep for N seconds mid-turn — if you're about to answer a lone "Hi", say so briefly and invite the real question rather than pretending time has passed. Reserve the literal timer for mode 2 (the @nohello/core engine), which is built for exactly this.

3. If the wait elapses with no follow-up, nudge

Render a reply in the configured tone. Keep it short, name the sender if known, and always end by inviting the actual question — never scold without inviting the ask.

• professional — polite, efficient, no jokes. "Hi Bala, thanks for reaching out — could you share what you need? Happy to help as soon as I have the full context."
• friendly — warm, casual, an emoji is fine. "Hey Bala 👋 go ahead and drop your question whenever, no need to wait for a reply first!"
• satirical — playful ribbing about the #nohello anti-pattern, still kind. "Achievement unlocked, Bala: 'Hello'. Now for the legendary 'Hello + Question' combo!"
• custom — use the operator-supplied template verbatim, substituting {name} and {waitSeconds}.

The full template sets live in packages/core/src/responses.ts.

Configuration

Same options in every integration (see packages/core/src/config.ts and docs/CONFIGURATION.md for full detail):

Option	Default	Meaning
enabled	true	Master on/off switch
waitSeconds	90	How long to wait for the real question before nudging
tone	"professional"	professional | friendly | satirical | custom
customTemplate	—	Required when tone is custom; supports {name}, {waitSeconds}
cooldownSeconds	600	Don't nudge the same sender in the same conversation again within this window
maxGreetingWords	8	Messages longer than this are never treated as greeting-only
maxLeftoverWords	2	Words allowed to remain (e.g. a name) after scrubbing greeting/small-talk phrases

Wiring this into a real bot

Don't reimplement the timer or templates by hand — install @nohello/core and call NoHelloEngine.ingestMessage() for every inbound message; it tells you when to schedule a nudge and calls your onNudge callback with the rendered text when the wait elapses uninterrupted:

import { NoHelloEngine } from '@nohello/core';

const engine = new NoHelloEngine({
  config: { waitSeconds: 60, tone: 'friendly' },
  onNudge: async (nudge) => {
    await postMessageToYourChatPlatform(nudge.conversationId, nudge.text);
  },
});

// on every inbound message from your chat platform's event handler:
await engine.ingestMessage({
  conversationId: event.channelId,
  userId: event.userId,
  userName: event.userDisplayName,
  text: event.text,
});

Reference adapters: packages/slack-adapter (Slack Bolt) and packages/teams-adapter (Microsoft Bot Framework). For Zoom, LinkedIn, or any other chat engine, see docs/INTEGRATIONS.md — the pattern is identical: forward inbound messages to ingestMessage, send whatever onNudge gives you back.