A useful agent should learn the work you keep giving it. Skill Workshop puts a review step in front of skill changes - proposals start as PROPOSAL.md before becoming SKILL.md. Shape your agent's reusable workflows before they become permanent behavior.
Every ClawHub skill now ships with a Skill Card documenting what the skill does and where it came from, scanned by SkillSpector for hidden instructions and other agentic risks. The ClawScan pipeline combines static analysis, VirusTotal, and NVIDIA SkillSpector.
OpenClaw is adding opt-in auto mode for Enterprise-ready host exec guardrails: policy runs first, low-risk misses get reviewed, and humans stay in the loop. Safe, repeatable commands can run without nagging you.
A release sweep across February through May shows faster agent turns, fewer dependencies, and a cleaner package shape. Performance optimizations and dependency cleanup across the codebase.
The security roadmap for making OpenClaw a powerful personal assistant runtime users can understand, observe, and trust. Covers transparency, auditability, and user control.
OpenClaw now runs OpenAI agent turns through the native Codex app-server harness by default, while bringing the lessons back to every model. Better integration with OpenAI's ecosystem.