Agent behavior manipulation via served plugin instructions
Published Jun 1, 2026 · Updated Jun 1, 2026
Prompt injection in Hermes Agent through 2026.4.30 allows remote attackers to alter agent behavior through plugin skill content. The _serve_plugin_skill() path detects known injection strings in SKILL.md but only logs a warning before skill_view() returns the instructions. Reachability requires an enabled plugin and selection of its skill content; the vendor classifies prompt injection and malicious plugin content as outside its security boundary.
Summary
What happened
Prompt injection in Hermes Agent through 2026.4.30 allows remote attackers to alter agent behavior through plugin skill content. The _serve_plugin_skill() path detects known injection strings in SKILL.md but only logs a warning before skill_view() returns the instructions. Reachability requires an enabled plugin and selection of its skill content; the vendor classifies prompt injection and malicious plugin content as outside its security boundary.
The record
- CVE
- CVE-2026-10220
- Published
- Jun 1, 2026
- Updated
- Jun 1, 2026
- Vendor
- Nous Research
- Product
- Hermes Agent
- Classifications
- CWE-74, CWE-707
- Attack vector
- network
- Privileges
- unauthenticated
Timeline
How it unfolded
- Jun 1, 2026CVE publishedPublication date reported by the CVE source.
- Jun 1, 2026Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=2026.4.0
- Affected versionversion=2026.4.1
- Affected versionversion=2026.4.10
- Affected versionversion=2026.4.11
- Affected versionversion=2026.4.12
- Affected versionversion=2026.4.13
- Affected versionversion=2026.4.14
- Affected versionversion=2026.4.15
- Affected versionversion=2026.4.16
- Affected versionversion=2026.4.17
- Affected versionversion=2026.4.18
- Affected versionversion=2026.4.19
- Affected versionversion=2026.4.2
- Affected versionversion=2026.4.20
- Affected versionversion=2026.4.21
- Affected versionversion=2026.4.22
- Affected versionversion=2026.4.23
- Affected versionversion=2026.4.24
- Affected versionversion=2026.4.25
- Affected versionversion=2026.4.26
- Affected versionversion=2026.4.27
- Affected versionversion=2026.4.28
- Affected versionversion=2026.4.29
- Affected versionversion=2026.4.3
- Affected versionversion=2026.4.30
- Affected versionversion=2026.4.4
- Affected versionversion=2026.4.5
- Affected versionversion=2026.4.6
- Affected versionversion=2026.4.7
- Affected versionversion=2026.4.8
- Affected versionversion=2026.4.9
What conditions does exploitation require?
What is affected?
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Weakness, pattern, technique
Public exploit references
- CVE-2026-10220 proof-of-concept gistproof of concept · unverified
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.
Book a demo