PatchSiren

PatchSiren cyber security CVE debrief

CVE-2026-10220 NousResearch CVE debrief

A medium-severity injection vulnerability in NousResearch hermes-agent, affecting versions up to 2026.4.30. The flaw resides in the _serve_plugin_skill/skill_view function within tools/skills_tool.py. Remote attackers can exploit this issue to perform injection attacks. The exploit has been publicly disclosed, and the vendor was contacted but did not respond.

Vendor
NousResearch
Product
hermes-agent
CVSS
MEDIUM 5.5
CISA KEV
Not listed in stored evidence
Original CVE published
2026-06-01
Original CVE updated
2026-06-01
Advisory published
2026-06-01
Advisory updated
2026-06-01

Who should care

Organizations running NousResearch hermes-agent instances, particularly those exposing skill/plugin functionality to remote or untrusted users. Security teams should prioritize review given public exploit availability and vendor non-response.

Technical summary

The vulnerability exists in the _serve_plugin_skill/skill_view function of tools/skills_tool.py in NousResearch hermes-agent through version 2026.4.30. Insufficient neutralization of special elements allows remote attackers to inject malicious input, leading to injection-based compromise. The attack vector is network-accessible with low attack complexity.

Defensive priority

medium

Recommended defensive actions

  • Review and restrict network access to hermes-agent instances where possible.
  • Inspect tools/skills_tool.py for the _serve_plugin_skill and skill_view functions; apply input validation and output encoding to mitigate injection risks.
  • Monitor for unauthorized or unexpected plugin skill execution in hermes-agent deployments.
  • Apply updates from NousResearch if and when a patched version is released.
  • Review the publicly disclosed technical details to assess exposure and implement targeted mitigations.

Evidence notes

Vuldb-assigned CVE with CVSS 4.0 vector. CWE-74 (Improper Neutralization of Special Elements in Output Used by a Downstream Component) and CWE-707 (Improper Neutralization) are cited as weakness types. The vendor field is marked low-confidence and needs review.

Official resources

Public disclosure occurred with vendor non-response. A public proof-of-concept is available.