CVE-2026-109331: Network-AI 5.14.0 before 5.15.2 Deny Pattern Bypass via ClaudeHookBridge extractFullTarget
Network-AI from 5.14.0 before 5.15.2 contains a security control bypass vulnerability in the ClaudeHookBridge PreToolUse hook, which matches deny patterns only against the first present candidate field. A prompt-injected agent can place dangerous payloads in fields like Write content or an MCP script parameter beside a benign filepath or command, evading denyPatterns to write web shells or execute scripts.
Affected Software
Event History
Frequently Asked Questions
Which deployments are affected?
Network-AI versions 5.14.0 through versions before 5.15.2 are affected. The issue concerns deployments using the ClaudeHookBridge PreToolUse hook and its denyPatterns control.
What does an attacker need to exploit this issue?
An attacker needs to prompt-inject an agent so it supplies a dangerous payload in a field that is not the first candidate field inspected. A benign file_path or command can be included alongside malicious Write content or an MCP script parameter to evade matching.
What should teams do to remediate the issue?
Upgrade Network-AI to version 5.15.2 or later. The provided data does not identify a workaround for affected versions.