Openclaw CVE-2026-35654

The CVSS v3.1 score of 5.3 reflects a medium-severity vulnerability with network attack vector (AV:N), low complexity (AC:L), and no privilege requirements (PR:N), indicating that any unauthenticated remote attacker can exploit this without special conditions. The impact is integrity-focused (I:L) with no confidentiality or availability loss, meaning attackers can modify or inject unauthorized feedback data rather than access or disrupt systems. While the CVSS score is moderate, the practical risk depends on the sensitivity of feedback data within organizational Teams deployments and the value of maintaining audit integrity for session feedback. The authentication bypass nature (CWE-288) elevates concern beyond a simple validation flaw, as it directly undermines trust in the feedback mechanism. The fact that a vendor patch is available and this was disclosed responsibly (GitHub security advisory) suggests coordinated disclosure rather than widespread active exploitation. In SSVC terms, this is exploitable with network access and low complexity, but the integrity-only impact and lack of evidence of widespread abuse suggest this is a medium priority requiring timely patching but not an immediate critical emergency.

An attacker can craft a network request to the OpenClaw feedback invoke endpoint without providing valid sender credentials, bypassing the allowlist check that should restrict feedback operations to authorized senders. The attacker then triggers unauthorized session feedback recording or reflection, potentially capturing or modifying feedback data intended for restricted participants. This could be chained with social engineering or internal reconnaissance to target specific Teams channels or sessions within an organization.