How I built a TypeScript CLI for Microsoft To Do in a single session with Copilot, packaged it with GitHub Actions, and integrated it as a shared OpenClaw skill so every agent can manage my task lists.
How I expanded my OpenClaw gateway from two agents to five, routed each to its own Teams channel, and discovered why RSC permissions matter more than config files.
Microsoft Agent Framework 1.0 just went GA. Here's what actually matters in it: stable APIs, A2A, declarative definitions, and what it still doesn't answer for you.
A small routing detail broke OpenClaw reminder delivery in exactly the wrong place. Here's how I traced it, fixed same-channel delivery, and stopped the agent from claiming success when scheduling had actually failed.
How I split OpenClaw into main and personal agents, gave them separate workspaces, and routed Signal and Teams differently so the system felt more natural to live with.
Why complex enabling work often disappears in large organizations, and why I now think of visibility as part of delivery rather than as a political side quest.
A walkthrough of the Dockerfile I use for OpenClaw: why I extend the base image, what I add for Signal and Teams, and why the order of a few lines matters more than it looks.
How I got OpenClaw talking to Signal and Microsoft Teams from one custom Docker image, and why the base image needed a few extra runtime pieces to make that practical.
The pattern I use to turn SonarQube from passive reporting into an active remediation loop with labels, issues, AI-assisted fixes, and human-controlled pull requests.
Why I chose to run OpenClaw in Docker instead of directly on the host: cleaner dependencies, repeatable rebuilds, durable state, and a setup that plays nicely with tunnels and self-hosted services.
