personal-assistant-stack

Design an approval-gated personal assistant stack for omnichannel messaging.

6|1|Updated May 13, 2026

Search Tags:#google calendar #ai assistant #approval workflow #openclaw #beeper #message triage #cost controls

Ready to run

npx skills add https://github.com/Xipher-Labs/walter-os --skill personal-assistant-stack

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: personal-assistant-stack
Source: https://github.com/Xipher-Labs/walter-os/tree/main/skills/personal-assistant-stack
Command: npx skills add https://github.com/Xipher-Labs/walter-os --skill personal-assistant-stack

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you design a personal AI assistant strategy that can safely handle inbound messages, surface what’s urgent, and optionally create calendar events from chat context without giving the assistant uncontrolled ability to send or modify things.

Core Features & Use Cases

Chat-first personal assistant routing: Uses OpenClaw as the bidirectional assistant so inbound messages are understood in the same context where replies are authorized.
Priority scoring and approval-based actions: Categorizes messages as urgent, can-wait, or ignore, with auto-template replies gated by per-contact opt-in and new/unusual replies requiring explicit approval.
Calendar event creation from messages: Extracts event details (e.g., “coffee at 5pm Thursday”) and inserts events via a scoped Google Service Account to limit blast radius.
Unified multi-channel inbox via Beeper: Uses Beeper as the transport layer to bridge WhatsApp/iMessage/Instagram/Signal and more into a single Matrix-compatible stream for OpenClaw.
Cost control with LiteLLM virtual keys: Routes routine vs reasoning-heavy tasks to different models with monthly caps, keeping automation costs bounded.
Safety boundaries for sensitive actions: Defines what is allowed by default (read/summary, confirmation for events) and what is never allowed (paying, logging into new accounts, firing unsolicited sends).

Quick Start

Tell the assistant to produce an approval-based personal assistant workflow using OpenClaw plus Beeper, including urgent/can-wait/ignore routing and a Google Calendar event extraction plan.

Frequently Asked Questions about personal-assistant-stack

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I build an AI assistant that can safely triage inbound messages across WhatsApp and Telegram?▼

To build a safe personal AI assistant, you can use a Beeper-bridged unified inbox to route omnichannel messages to OpenClaw. The assistant then triages inbound messages by categorizing them as urgent, can-wait, or ignore before drafting any replies.

How does approval-gated message sending work for an automated personal assistant?▼

Approval-gated sending requires explicit user authorization for new or unusual replies, while auto-template replies are gated by per-contact opt-in. This mechanism ensures the AI assistant never fires unsolicited sends without your direct consent.

Can I extract Google Calendar events directly from chat context using an AI assistant?▼

Yes, you can extract calendar events from chat context by having the assistant identify details like meeting times, then inserting them via a scoped Google Service Account. This limits the blast radius by restricting the assistant's calendar permissions.

What is the best way to control API costs when automating message triage and reply generation?▼

The best way to control automation costs is using LiteLLM virtual-key caps to route routine tasks to cheaper models and reasoning-heavy tasks to advanced ones. This applies monthly spending caps to keep LLM costs bounded.

What are the safety boundaries when using an AI assistant to handle omnichannel messaging?▼

Safety boundaries define what is allowed by default, such as reading or summarizing messages and confirming events. The assistant is never allowed to perform sensitive actions like paying, logging into new accounts, or firing unsolicited sends.

Does Beeper work with OpenClaw for unifying multi-channel inboxes?▼

Yes, Beeper works with OpenClaw by acting as the transport layer to bridge WhatsApp, iMessage, Instagram, and Signal into a single Matrix-compatible stream. This allows the assistant to process all omnichannel messaging in one unified context.