CVE-2026-130121: AI Puffer <= 2.5.28 - Unauthenticated Stored Cross-Site Scripting via AI-Generated Comment Replies
The AI Puffer - MCP Server for Claude & ChatGPT, with AI Chatbot plugin for WordPress is vulnerable to Stored Cross-Site Scripting via AI-generated comment replies in all versions up to, and including, 2.5.28 due to insufficient sanitization of the AI model's output in the insertcommentreplylogic() function, which saves replies with wpinsertcomment() and bypasses the kses filtering WordPress normally applies to comment content. This makes it possible for unauthenticated attackers to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page, by submitting a comment that leads the AI model to reproduce an HTML payload in its automated reply. This is only exploitable when an administrator has configured a Comment Auto-Reply automation task with 'Action on reply' set to 'Approve' (the default is 'Hold for moderation'), and the attacker's comment has been approved.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Configuration
Set 'Action on reply' to 'Hold for moderation' instead of 'Approve'.
AI Puffer Comment Auto-Reply automation task Action on reply = Hold for moderation
Event History
Frequently Asked Questions
Is the default Comment Auto-Reply configuration affected?
No. Exploitation requires the Comment Auto-Reply task to have “Action on reply” set to “Approve”; the default setting is “Hold for moderation.”
What conditions must an attacker meet to exploit this issue?
The attacker must submit a comment that causes the AI model to reproduce an HTML payload in its automated reply. The attacker’s original comment must also be approved, and the automation must be configured to approve replies automatically.
Who is exposed to the injected script?
Any user who accesses a page containing the injected AI-generated comment reply may execute the attacker’s script in their browser. The vulnerability allows unauthenticated attackers to supply the triggering comment.
What can be done before an update is available?
Set the Comment Auto-Reply task’s “Action on reply” setting to “Hold for moderation” rather than “Approve.” Review and carefully moderate comments that could trigger automated AI replies before approving them.
How can administrators check for potential exploitation?
Review pages and approved comment threads containing AI-generated replies for unexpected HTML or script-like content. Pay particular attention to replies created while the automation was configured to approve replies automatically.