Remote Command Injection in 0x4m4 HexStrike AI
A command injection vulnerability in HexStrike AI allows remote unauthenticated attackers to execute arbitrary OS commands via the Execute Endpoint.
CVE search metadata
CVE search record: CVE-2026-90619. Severity: high. CVSS: 7.3. KEV: no. Product: HexStrike AI (<= d689933ff579d839c676c82b231f8e98326c5f04), HexStrike AI (up to commit d689933ff579d839c676c82b231f8e98326c5f04). Brief: Remote Command Injection in 0x4m4 HexStrike AI. Brief link: https://feed.craftedsignal.io/briefs/2026-09-hexstrike-rce/
What's new
- 1. added coverage for HexStrike AI (up to commit d689933ff579d839c676c82b231f8e98326c5f04) Sep 14, 03:30 via nvd
HexStrike AI, an open-source project by 0x4m4, contains a remote code execution vulnerability (CVE-2026-90619) affecting all releases up to commit d689933ff579d839c676c82b231f8e98326c5f04. The flaw resides within the 'Execute Endpoint' component inside the 'hexstrike_server.py' file. An attacker can trigger this vulnerability by sending a maliciously crafted request to the application, specifically targeting the 'code' or 'script' arguments. Because the input is processed without adequate sanitization, the application passes the user-supplied data directly to the underlying operating system's shell, resulting in arbitrary command execution. This vulnerability is remotely exploitable without authentication, and public proof-of-concept exploits exist, posing a high risk to organizations utilizing this component in production environments. As the project follows a continuous delivery model without versioned releases, users must monitor the project repository for updates.
Impact
The vulnerability allows full remote code execution on the server hosting the HexStrike AI component. Successful exploitation leads to total system compromise, including potential data exfiltration, deployment of malicious payloads, or use of the server as a pivot point within the network. Because the vulnerability is remotely reachable and requires no authentication, it is highly attractive for opportunistic exploitation across any publicly exposed instances.
Recommendation
- Identify all instances of the HexStrike AI project within the environment and restrict network access to the 'Execute Endpoint' component until a patch is available.
- Implement an application firewall or proxy-level filter to inspect incoming HTTP requests for suspicious shell metacharacters in the 'code' or 'script' query parameters.
- Monitor webserver logs for requests directed at the Execute Endpoint containing patterns indicative of command injection (e.g., semicolons, pipe operators, or backticks).
- Monitor process creation events on servers running HexStrike AI for unusual child processes (e.g., cmd.exe, /bin/sh, nc) spawned by the server process.
Immediate actions
Deploy webserver detection rule to identify attempted exploitation
Mitigations
Restrict external network access to the HexStrike AI Execute Endpoint
CVE-2026-90619
Detection coverage 1
Detects CVE-2026-90619 Exploitation - Remote OS Command Injection
highDetects HTTP requests to HexStrike AI containing shell metacharacters in the code or script arguments.
Detection queries are available on the platform. Get full rules →