Openclaw EUVD-2026-12722

The CVSS v3.1 vector (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:L) indicates a network-accessible vulnerability requiring no authentication or user interaction, but with relatively low complexity to exploit. The score of 6.5 reflects impacts on both integrity (unintended content removal via regex manipulation) and availability (denial of service via backtracking). The availability of two confirmed patches via GitHub commits (7e67ab75cc2f0e93569d12fecd1411c2961fcc8c and 74268489137510b6f6349919d1e197b17290d92c) demonstrates vendor responsiveness and confirms the technical validity of the issue. The ENISA EUVD tracking (EUVD-2026-12722) and VulnCheck advisory corroborate the finding, and the public GitHub security advisory indicates some level of disclosure. While KEV status is not explicitly provided in the intelligence, the combination of publicly disclosed patches, network-accessible attack vector, and lack of required privileges elevates real-world risk for OpenClaw deployments exposed to untrusted Feishu message sources. This is a legitimate priority for organizations running affected versions, particularly those integrating Feishu as a user-facing messaging channel.

An attacker registers or compromises a Feishu account within an organization's workspace and sends a message containing a maliciously crafted mention with regex metacharacters such as @user<regex(a+)+b|nested-quantifier-pattern> to an OpenClaw bot. When the stripBotMention function attempts to construct a RegExp object from this unescaped metadata without input validation, the regex engine enters catastrophic backtracking, consuming CPU resources and preventing the bot from processing any subsequent messages for minutes at a time, effectively blocking the bot's availability. Alternatively, an attacker could craft mention patterns designed to match and strip legitimate content before the bot's natural language model processes the message, allowing content injection or manipulation attacks.