Back to projects

OpenClaw

Three-Tier Agent Stack

My personal AI operating system — self-configured Skills across 7 domains, with Minimax-powered Nova triaging daily ops, HealthBot running my training plan, and Claude Code handling strategy and code. Built and run by me, every day.

Type
Personal AI Infrastructure
Period
2026.04 — Present (continuously evolving)
Team
Solo (PM + Prompt Engineer + Architect)
Outcome
Daily-driven · continuously evolving since 2026.04
Stack
OpenClaw + Claude Code + MiniMax M2 + Cron / Webhook / WeChat Push
My Role
ArchitectPrompt EngineerOperatorDaily User

English summary

Translation in progress

A three-tier personal AI ops stack: Nova (Minimax-powered secretary) triages daily tasks for cheap tokens, HealthBot runs my training plan via WeChat push, and Claude Code (Opus 4.7) handles strategic decisions and code. 104 self-configured OpenClaw Skills across 7 domains: content, health, career, finance, learning, automation, monitoring. Designed as a 'one-person company command center' — the clearest evidence of moving from AI tool user to tool-chain builder.

Chinese version below — covers the architectural rationale (cheap-tokens-for-ops + expensive-tokens-for-decisions), the seven operational domains, key skill examples, and what running it for a year taught me about agent design.

Back to all projects