@wide-doctor-37830 review comments are all addressed on PR #2472 (
https://github.com/OpenHands/software-agent-sdk/pull/2472). After it lands, I have three follow-up issues queued from a red-team review: (1) per-field extraction cap to prevent truncation weaponization, (2) two-phase rm flag detection to close flag-ordering bypasses, and (3) expanded pattern coverage for reverse shells, privilege escalation, equivalent-destruction commands, process substitution, and base64-decode-to-exec.
I don't want to expand scope on the current PR, but also can't responsibly let these vulns go untracked. I can draft the follow up issues if desired, so the follow-ups can be cascade merged clean and quick.