OpenClaw Foundation

What happens when an AI agent’s permissions change while it’s still working? That was one of the security questions explored in OpenClaw’s recent audit with Trail of Bits through OpenAI's Patch the Planet initiative. The review examined how permissions, identity, and approvals hold up as agents carry out work across multiple steps. Every actionable issue has been repaired, with fixes shipped in stable releases. 🤖 Josh Avant's recap explains what the team found, what changed, and what we learned about securing agent systems. https://lnkd.in/eAp7NaBa

OpenClaw v2026.9.4 is out 🦞 OpenClaw 2026.9.4 makes plugins and skills easier to find, and lets you turn past conversations into reusable skills through a chat you can steer. It also adds GPT Image 2.5 support, more control over cloud sessions, and interactive questions in the terminal, alongside fixes for updates, memory, and messaging. In Skill Workshop, "Learn from past conversations" opens a conversation you can follow, give direction to, or stop. Choose Propose to review suggested improvements before applying them, or Auto to let your Claw apply improvements as it works. Starting that chat does not turn on automatic self-learning. Finding the next useful capability is easier too. You can search installed skills alongside ClawHub, browse installed and available plugins together, and review a plugin before installing it. This release brings together 1,558 pull requests and 20 direct commits from 293 contributors. Thank you to everyone who helped build, review, test, and improve it. Full release notes https://lnkd.in/gPaPBs5J