Systems engineer building the infrastructure layer for autonomous agents. Interested in agents that are inspectable, recoverable, and don't turn into black boxes you have to trust blindly.
Prefers file-backed state over invisible prompt context. Believes verification should be mandatory, not optional. Builds with the human in the loop — always.
Open source, open thinking. Most of what I work on lives on GitHub.
Inspectable autonomous agent loop scaffold. Observe → Condense → Plan → Execute → Verify → Persist. File-backed state means every decision, every action, and every result survives restarts and is always auditable. Multiple executors: research, code, browser, messaging.
Chrome DevTools with accessibility-tree semantics. Use your authenticated Chrome sessions with element targeting via accessibility refs instead of fragile CSS selectors. Auto-reconnect built in.
These are the AI capabilities running on my behalf — built to be specific, reliable, and genuinely useful rather than novelty demos.
Drop a YouTube URL and get a real summary plus analytical depth — not just auto-generated captions. Extracts the full transcript first, then synthesises key claims, implications, credibility, and connections to other topics.
git clone https://github.com/sidonsoft/youtube-analysis-skill.git ~/.openclaw/skills/youtube-analysis
If you want to talk about open source, agentic systems, or have something interesting to build — reach out.