PatchSiren cyber security CVE debrief
CVE-2026-9367 NousResearch CVE debrief
A command injection vulnerability exists in NousResearch hermes-agent, affecting the `detect_dangerous_command` function within `tools/approval.py` of the terminal_tool component. The flaw permits remote attackers to inject arbitrary operating system commands. The vulnerability was publicly disclosed on 2026-05-24 after the vendor was contacted but did not respond. A proof-of-concept exploit has been publicly released. The vulnerability is classified as MEDIUM severity with a CVSS score of 5.5.
- Vendor
- NousResearch
- Product
- hermes-agent
- CVSS
- MEDIUM 5.5
- CISA KEV
- Not listed in stored evidence
- Original CVE published
- 2026-05-24
- Original CVE updated
- 2026-05-26
- Advisory published
- 2026-05-24
- Advisory updated
- 2026-05-26
Who should care
Organizations running NousResearch hermes-agent with terminal_tool enabled, particularly those exposing agent interfaces to untrusted networks or integrating with automated workflow systems.
Technical summary
The vulnerability resides in the `detect_dangerous_command` function within `tools/approval.py` of the terminal_tool component in NousResearch hermes-agent. Insufficient input sanitization allows remote attackers to inject and execute arbitrary operating system commands. The attack vector is network-accessible with low attack complexity. The affected code is present in commits up to and including 5157f5427f19488b31c6fdebbacd15d798ce7f63.
Defensive priority
medium
Recommended defensive actions
- Review and restrict access to hermes-agent deployments, particularly those exposing terminal_tool functionality
- Audit `tools/approval.py` for unsafe command execution patterns and implement strict input validation
- Apply principle of least privilege to any service accounts used by hermes-agent
- Monitor for suspicious command execution patterns in environments running affected versions
- Subscribe to vendor security advisories for NousResearch/hermes-agent repository for patch availability
Evidence notes
Vulnerability affects commit 5157f5427f19488b31c6fdebbacd15d798ce7f63 and earlier. The `detect_dangerous_command` function in `tools/approval.py` fails to properly sanitize input, enabling OS command injection. CWE-77 and CWE-78 are identified as relevant weakness classifications. The exploit has been publicly disclosed via GitHub Gist.
Official resources
public