OpenClaw — Personal AI Automation System

Problem

Each AI tool I used carried its own context, its own workflow, its own switching cost. Nothing held state between them. I wanted one operating model where the assistant kept context across sessions and ran inside my real workflows, not a separate window I had to load up every time.

What I Built

OpenClaw runs on Claude Code’s plugin and skill architecture:

• Skills: Reusable instructions for recurring work, including career OS operations, code review patterns, content production, and session management.
• MCPs (Model Context Protocol servers): Integrations that connect the assistant to my Obsidian vault, calendar, and file system, so it reads my actual context instead of an isolated prompt.
• Workflow automation: I express recurring jobs, including CV tailoring, interview prep, project documentation, and research synthesis, as skills that call the right tools in the right order.
• Memory system: File-based memory that carries context from one session into the next.

Operating Model

I use OpenClaw as a daily operating system for technical work: drafting and maintaining career materials, exploring codebases, designing and implementing features, synthesizing research, and managing the Obsidian vault that backs this portfolio.

Outcome

I move faster across development, writing, and knowledge management. When I improve one skill, every workflow that calls it gets the benefit, so I keep refining the ones I run most.