OpenBao Agent’s exec rendering mode can write secret values from env_template to standard output after repeated rendering failures. When the template runner is recreated, the agent emits the rendered secrets, which any entity that can read the agent’s stdout can capture. This data‑leak exposes confidential credentials or tokens, violating confidentiality.

The vulnerability affects OpenBao Agent deployments prior to version 2.6.0. Versions 2.6.0 and later contain the fix and are not impacted.

The CVSS base score is 2.4, indicating a low severity, and the vulnerability is not listed in the CISA KEV catalog. Because it exposes data only through the agent’s standard output, the attack vector is likely local or through a process that runs with the same privileges as the agent. An attacker with such local access, or a supervisor that logs or captures all agent output, could capture the rendered secrets. Although the exploitability is low, the potential exposure of secret data remains a concern until the fix is applied.