Skills

MCP, OpenAPI tool surfaces, sandboxed execution, human approval, evals, and runtime ergonomics for agents that do real work.

Predictive AI, generative AI, governance, explainability, observability, permissions, productization, and production reliability.

Role-separated agents, PR babysitting, CI, browser QA, smoke tests, rollback, approval queues, and agent-led delivery systems.

Codex, Claude Code, OpenClaw, skills, slash commands, PR review loops, CI, and workflow automation for agent-led development.

Swoleby, behavior loops, SMS interfaces, user state, reminders, subscriptions, evals, and product judgment under real constraints.