home / skills / openclaw / skills / smart-followups

smart-followups skill

/skills/robbyczgw-cla/smart-followups

This skill suggests contextual follow-up questions after AI responses to keep conversations flowing and deepen understanding.

npx playbooks add skill openclaw/skills --skill smart-followups

Review the files below or copy the command above to add this skill to your agents.

Files (20)
SKILL.md
3.5 KB
---
name: smart-followups
description: Generate contextual follow-up suggestions after AI responses. Shows 3 clickable buttons (Quick, Deep Dive, Related) when user asks for "followups".
triggers:
  - followups
  - follow-ups
  - suggestions
  - give me suggestions
  - what should I ask
channels:
  - telegram
  - discord
  - slack
  - signal
  - whatsapp
  - imessage
  - sms
  - matrix
  - email
---

# Smart Follow-ups Skill

Generate contextual follow-up suggestions for Clawdbot conversations.

## How to Trigger

Say any of these to get follow-up suggestions:

| Trigger | Example |
|---------|---------|
| `followups` | "followups" |
| `follow-ups` | "give me follow-ups" |
| `suggestions` | "any suggestions?" |
| `what next` | "what should I ask next?" |

> **Note:** This is a keyword the agent recognizes, not a registered `/slash` command. Clawdbot skills are guidance docs that tell the agent how to respond.

## Usage

Say "followups" in any conversation:

```
You: What is Docker?
Bot: Docker is a containerization platform...

You: /followups

Bot: 💡 What would you like to explore next?
[âš¡ How do I install Docker?]
[🧠 Explain container architecture]
[🔗 Docker vs Kubernetes?]
```

**On button channels (Telegram/Discord/Slack):** Tap a button to ask that question.

**On text channels (Signal/WhatsApp/iMessage/SMS):** Reply with 1, 2, or 3.

## Categories

Each generation produces 3 suggestions:

| Category | Emoji | Purpose |
|----------|-------|---------|
| **Quick** | âš¡ | Clarifications, definitions, immediate next steps |
| **Deep Dive** | 🧠 | Technical depth, advanced concepts, thorough exploration |
| **Related** | 🔗 | Connected topics, broader context, alternatives |

## Authentication

**Default:** Uses Clawdbot's existing auth — same login and model as your current chat.

**Optional providers:**
- `openrouter` — Requires `OPENROUTER_API_KEY`
- `anthropic` — Requires `ANTHROPIC_API_KEY`

## Configuration

```json
{
  "skills": {
    "smart-followups": {
      "enabled": true,
      "provider": "clawdbot",
      "model": null
    }
  }
}
```

| Option | Default | Description |
|--------|---------|-------------|
| `provider` | `"clawdbot"` | Auth provider: `clawdbot`, `openrouter`, `anthropic` |
| `model` | `null` | Model override (null = inherit from session) |
| `apiKey` | — | API key for non-clawdbot providers |

## Channel Support

| Channel | Mode | Interaction |
|---------|------|-------------|
| Telegram | Buttons | Tap to ask |
| Discord | Buttons | Click to ask |
| Slack | Buttons | Click to ask |
| Signal | Text | Reply 1-3 |
| WhatsApp | Text | Reply 1-3 |
| iMessage | Text | Reply 1-3 |
| SMS | Text | Reply 1-3 |
| Matrix | Text | Reply 1-3 |
| Email | Text | Reply with number |

See [CHANNELS.md](CHANNELS.md) for detailed channel documentation.

## How It Works

1. User types `/followups`
2. Handler captures recent conversation context
3. Clawdbot generates 3 contextual questions (using current model/auth)
4. Formatted as buttons or text based on channel
5. User clicks button or replies with number
6. Clawdbot answers that question

## Files

| File | Purpose |
|------|---------|
| `handler.js` | Command handler and channel formatting |
| `cli/followups-cli.js` | Standalone CLI for testing/scripting |
| `README.md` | Full documentation |
| `CHANNELS.md` | Channel-specific guide |
| `FAQ.md` | Common questions |

## Credits

Inspired by [Chameleon AI Chat](https://github.com/robbyczgw-cla/Chameleon-AI-Chat)'s smart follow-up feature.

Overview

This skill generates three contextual follow-up suggestions after an AI response and presents them as clickable buttons or simple numbered text depending on the channel. It helps steer conversations with concise next-step prompts in three categories: Quick, Deep Dive, and Related. The goal is to increase engagement and make it easy for users to continue the thread with useful, context-aware questions. It works with the current session authentication and can optionally use third-party API keys.

How this skill works

When a user requests follow-ups (for example by saying "followups" or "suggestions"), the skill captures recent conversation context and asks the active model to produce three tailored prompts. It formats those prompts as buttons on supported platforms (Telegram, Discord, Slack) or as numbered text the user can reply to on text-only channels. Tapping a button or replying with a number sends the selected prompt back to the agent, which then answers using the same session model/authentication.

When to use it

  • When the user asks "followups", "suggestions" or "what next"
  • After a high-level answer to suggest immediate clarifications
  • When a technical topic needs deeper exploration
  • To link the current topic to related concepts or alternatives
  • In chat flows where quick user action is preferred (buttons/text replies)

Best practices

  • Trigger follow-ups only after a substantive response to preserve relevance
  • Prefer short, actionable prompts for Quick suggestions and detailed framing for Deep Dive
  • Honor channel capabilities: use buttons where supported and numbered replies on text channels
  • Keep the three suggestions complementary (clarify, expand, relate) to cover common user intents
  • Incorporate session model and auth so suggestions match the agent's knowledge and tone

Example use cases

  • After a definition or overview (e.g., "What is Docker?") offer install steps, architecture deep dive, and comparison topics
  • In troubleshooting: suggest quick checks, a deep diagnostic path, and related configuration topics
  • During learning sessions: propose a simple quiz question, an advanced theory explanation, and related technologies to explore
  • In product discovery: propose a feature quick summary, deeper implementation details, and competitor comparisons

FAQ

What triggers the follow-ups?

Typing keywords like "followups", "follow-ups", "suggestions" or asking "what next" triggers the generation.

How do buttons vs text replies work?

Platforms with button support show three clickable options; text-only channels display numbered suggestions the user can reply with (1, 2, or 3).

Which model and authentication are used?

By default the skill uses the current session provider and model. You can override provider or supply API keys for supported external providers.