OpenClaw CVE-2026-41363

Exploitation requires: (1) valid authentication credentials to access OpenClaw and invoke the Feishu extension, (2) the Feishu extension must be enabled in the target deployment, (3) the upload_image operation must be accessible and functional, (4) the target file must be readable by the OpenClaw process (within its filesystem permissions). Non-default configuration is not an explicit prerequisite - the vulnerability exists in standard enabled-extension deployments. The AC:H (high complexity) rating suggests that successful exploitation may require specific path formats, knowledge of target file locations, or multiple attempts to bypass residual validation.

CVSS 6.0 with AC:H (high complexity) and PR:L (authenticated) indicates moderate real-world risk. The high-complexity attack vector suggests exploitation requires specific conditions or multiple steps, limiting casual exploitation. The PR:L requirement means only authenticated users can trigger the vulnerability, reducing the threat surface compared to unauthenticated attacks. However, VC:H (high confidentiality impact) confirms arbitrary file read capability affecting sensitive data. The KEV status is not provided in input data, and no public exploit code confirmation is stated, suggesting this is not yet in widespread active exploitation despite public disclosure via GitHub advisory. EPSS score is not provided but would likely be moderate given the authentication requirement and high complexity. Organizations with restrictive access controls limiting Feishu extension users will have lower real-world risk than those granting broad authentication to this feature.

An authenticated user with access to OpenClaw's Feishu extension crafts an upload_image request with a malicious path parameter containing traversal sequences (e.g., '../../etc/passwd' or an absolute path). The resolveUploadInput function fails to properly validate this path, allowing the application to read and return the contents of files outside the intended sandbox directory. The attacker obtains sensitive configuration files, credentials, or application source code accessible to the OpenClaw process.