Openclaw CVE-2026-35648

The CVSS score of 3.7 reflects low real-world severity despite the authentication-bypass classification tag. The vector (AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N) indicates network-accessible attack surface with high complexity and no authentication barrier, but limited impact (integrity only, no confidentiality or availability loss). The high attack complexity requirement suggests that successful exploitation depends on specific conditions: timing the attack with policy transitions, understanding the queuing mechanism, and having actionable commands already pending. No public exploit code has been identified, no active exploitation is confirmed via CISA KEV, and EPSS probability of exploitation is not provided. The practical risk is constrained to environments where command policies are frequently adjusted and operators rely on policy changes as a rapid access revocation mechanism. This is a legitimate but specialized threat requiring defenders to implement queue flushing or revalidation on policy changes rather than a systemic flaw affecting all installations.

An administrator configures OpenClaw to queue a batch of node commands for delayed execution, then later revokes access to sensitive operations and tightens the command policy. The attacker exploits the time window between policy restriction and command delivery: the queued commands reference an allowlist or policy declaration that was valid when queued but is now revoked. Because the system does not revalidate at delivery time, the commands execute despite the new policy, allowing an attacker to perform an action they should no longer have authorization for. Successful exploitation requires knowledge of pending queued actions and precise timing around policy changes, explaining the high attack complexity rating. No public proof-of-concept or active exploitation in the wild has been confirmed.