Executive brief
OpenBao Agent is a credential injection tool that authenticates to a secrets vault and injects credentials into child processes. Under certain error conditions when retry limits are exceeded, the Agent's exec rendering mode incorrectly writes sensitive secrets configured in environment templates to stdout, where they may be captured by logs or monitoring systems. This could expose credentials to unauthorized personnel with access to application logs.
Technical details
This is an information disclosure vulnerability (CWE-532: Insertion of Sensitive Information into Log File) in OpenBao Agent's exec rendering mode. When configured with `env_template` settings and error conditions trigger retry limit exhaustion (`num_retries`), the Agent writes unencrypted secret values to stdout instead of securely injecting them into the child process environment. The vulnerability requires local access to the machine running OpenBao Agent, high privileges on that system, and active user interaction to trigger the error condition. An attacker with these preconditions can capture credential material from stdout/logs. The fix is available in OpenBao v2.6.0 GA and later.
Affected products
- OpenBao Agent <2.6.0
Timeline
- 2026-07-14: disclosed
- 2026-09-22: advisory
- 2026-09-22: patched: Fixed in OpenBao v2.6.0 GA