CVE-2026-9354: Arbitrary Mass Mention Bypass in NousResearch hermes-agent Slack and Mattermost Adapters
Vulnerability ID: CVE-2026-9354
CVSS Score: 6.9
Published: 2026-05-24
A vulnerability in the Slack and Mattermost platform adapters for NousResearch hermes-agent permits an unauthenticated remote attacker to execute arbitrary mass mentions. By leveraging prompt injection, an attacker can bypass output sanitization logic and trigger workspace-wide notification exhaustion.
TL;DR
Unauthenticated prompt injection in hermes-agent Slack and Mattermost adapters allows attackers to trigger mass ping notifications via unescaped tags.
⚠️ Exploit Status: POC
Technical Details
- CWE ID: CWE-116 / CWE-20
- Attack Vector: Network (Prompt Injection)
- CVSS 4.0: 6.9 (Medium)
- CVSS 3.1: 5.4 (Medium)
- Impact: Availability (Notification Exhaustion)
- Exploit Status: Public PoC Available
- CISA KEV: Not Listed
Affected Systems
- NousResearch hermes-agent <= 2026.4.16 (Slack Adapter)
- NousResearch hermes-agent <= 2026.4.16 (Mattermost Adapter)
- hermes-agent: <= 2026.4.16
Exploit Details
- GitHub Gist: Proof of Concept reproducing the mass ping bypass against Slack and Mattermost endpoints
Mitigation Strategies
- Modify Slack adapter regex to exclude broadcast tags or implement secondary escaping pass
- Inject 'disable_mentions': true into Mattermost adapter payload props
- Implement universal LLM output guardrails to neutralize platform-specific mention syntax
Remediation Steps:
- Locate gateway/platforms/mattermost.py in the hermes-agent source
- Update the post request payload to include 'props': {'disable_mentions': True}
- Locate gateway/platforms/slack.py in the hermes-agent source
- Add a pre-transmission replace function for '<!everyone>', '<!channel>', and '<!here>'
- Restart the hermes-agent service to apply source modifications
References
- VulDB Entry for CVE-2026-9354
- CVE.org Record for CVE-2026-9354
- PoC Exploit Report
- hermes-agent Source Code Repository
Read the full report for CVE-2026-9354 on our website for more details including interactive diagrams and full exploit analysis.
Top comments (0)