personal-assistant-stacklisted
Install: claude install-skill Xipher-Labs/walter-os
# Walter-OS personal assistant stack — the answer is one assistant, not three
## The core decision
Three Telegram bots is the wrong shape. The right shape:
| Bot | Role | Surface |
|---|---|---|
| `@${WALTER_TELEGRAM_BOT_HANDLE}` | Outbound alerts only | Walter-VM monitors → push to me |
| `@${WALTER_OPENCLAW_BOT_HANDLE}` | Personal assistant — bidirectional | I talk to it, it reads my messages, it replies for me |
**No third bot.** OpenClaw IS the personal assistant. Adding a third loses the
context boundary (the assistant can't see what the alerter sent without a
relay) and adds another auth surface to manage.
## What OpenClaw covers
OpenClaw (openclaw.ai, self-hosted, Node 24+) connects to ALL of:
- WhatsApp (via Beeper or direct WA Business API)
- Telegram (the bot for outbound, plus your user account for read-only inbox)
- Slack, Discord, Google Chat
- iMessage, Signal (via Beeper bridge)
- Microsoft Teams, Matrix, Mattermost
- Instagram DMs (via Beeper bridge)
It's **bidirectional**: you message it, it reads on your behalf, it replies
when you authorize, it summarizes when you ask, it pings you on what matters.
## How it answers your specific use cases
### "Reply to messages for me when I'm busy"
Pattern: **pairing + pre-approved auto-reply rules**.
- New WhatsApp arrives → OpenClaw scores priority (urgent / can-wait / ignore)
- For "can-wait": auto-reply "Hey, I got your message and will answer in X hours"
using a template you defined for the contact's lo