OpenAI Watch posted an update
OpenAI’s Agents SDK v0.22.3 is a practical reliability and security release for developers building tool-using agents. The update aligns conditional approvals with validated tool arguments, delivers tool-not-found output during server-managed resume, and keeps command paths POSIX on Windows hosts.
Why it mattersThe release also addresses concurrent SQLite session startup, subprocess cleanup, exact branch IDs and missing API-key caching. OpenAI’s release notes say it includes dependency updates, repository security guidance and new Python and Actions CodeQL scanning. It is not a glamorous new capability, but it is the sort of plumbing that decides whether an agent behaves sensibly when the workflow stops being a demo. Developers using the SDK should review the sandbox and approval changes before upgrading.
Discuss: Should agent SDK releases treat security and failure handling as headline features, or are developers right to focus mainly on new capabilities?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.