nohello

A configurable #nohello skill: detects chat messages that are just a greeting ("hi", "hello", "hello <name>", "hi, how are you") with no actual question attached, waits a configurable number of seconds for the sender's real ask, and — if it never shows up — sends a configurable nudge back (professional, friendly, satirical, or your own custom tone).

Ships two ways:

• skills/nohello — a Claude Agent Skill so Claude (in Claude Code, Claude.ai, or a Claude-powered bot) applies the same detection and tone conventions inline.
• @nohello/core — a small, platform-agnostic TypeScript engine, with reference adapters for Slack and Microsoft Teams, so any chat engine can run this as a real, timer-driven bot.

Background on the actual etiquette this implements: nohello.org.

Packages

Package	Description
@nohello/core	Detection, wait-timer, and tone-template engine. No platform dependencies.
@nohello/slack	Slack (Bolt) adapter, plus a ready-to-run standalone bot.
@nohello/teams	Microsoft Teams (Bot Framework) adapter, plus a ready-to-run standalone bot.
examples/generic-webhook	Minimal HTTP-in/HTTP-out reference for wiring up any other chat engine (Zoom, internal tools, etc.).
skills/nohello	Claude Agent Skill packaging of the same behavior.

See docs/INTEGRATIONS.md for platform-specific setup (Slack, Teams, Zoom Team Chat, and an honest note on LinkedIn's API limitations), docs/CONFIGURATION.md for the full config reference, including per-team/per-user overrides and scaling beyond a single process, and docs/PUBLISHING.md for how CI and npm releases are wired up via GitHub Actions.

Quickstart

npm install
npm run build
npm test

Try it in 30 seconds (no chat platform needed)

NOHELLO_WAIT_SECONDS=5 NOHELLO_TONE=satirical npm run start --workspace nohello-example-terminal-chat

Type a bare hi, wait 5 seconds, get nudged. Type a real question before the timer elapses and the nudge is cancelled. See examples/terminal-chat.

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

const engine = new NoHelloEngine({
  config: {
    waitSeconds: 90,          // how long to wait for the real question
    tone: 'friendly',         // 'professional' | 'friendly' | 'satirical' | 'custom'
    cooldownSeconds: 600,     // don't re-nudge the same person too often
  },
  onNudge: async (nudge) => {
    // send nudge.text into your chat platform, addressed to nudge.conversationId
  },
});

// call this for every inbound message from your chat platform
await engine.ingestMessage({
  conversationId: 'channel-123',
  userId: 'user-456',
  userName: 'Bala',
  text: 'hi',
});
// if no follow-up arrives within waitSeconds, onNudge fires with a rendered nudge

How detection works

A message counts as "greeting-only" if, after stripping a greeting opener (hi, hello, hey, good morning, ...) and small-talk filler (how are you, you there?, quick question, ...), no more than a couple of words are left over (e.g. just a name). A message with real content attached — "Hi, can you review PR #123 today?" — is left alone. Full algorithm in packages/core/src/detector.ts; design rationale in docs/ARCHITECTURE.md.

Development

This is an npm workspaces monorepo (Node 18+).

npm install                 # install all workspace deps
npm run build --workspaces  # build every package
npm run test --workspaces   # run @nohello/core's test suite (vitest)
npm run typecheck --workspaces

License

MIT © Balajikumar Murugan