# CVE-2026-102874: HKUDS AnyTool command injection in Execute Endpoint

Source: https://junglewise.ai/threats/cve-2026-102874-a-vulnerability-was-identified-in-hkuds-anytool-0-1-0-affected  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-09-30. Updated: 2026-09-30  
License: CC BY 4.0 with a link back

CVE-2026-102874 · Severity: high · CVSS 7.3 · Published 2026-09-30

AnyTool is a universal tool-use layer for AI agents that exposes a local server with endpoints for executing commands. An unauthenticated attacker on the network can send a malicious request to the /execute endpoint to execute arbitrary OS commands with the privileges of the server process, reading sensitive data or compromising the system.

Vendors: [HKUDS](https://junglewise.ai/threats/vendors/hkuds).

## Quick answers

**Is CVE-2026-102874 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-102874 being exploited in the wild.

**How severe is CVE-2026-102874?** High severity, with a CVSS score of 7.3, published 30 September 2026.

**Which products and versions are affected?** HKUDS AnyTool 0.1.0.

**Is there a fix for CVE-2026-102874?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is CVE-2026-102874 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-102874?** Check whether you run HKUDS AnyTool at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## Technical details

The vulnerability is OS command injection (CWE-78) in anytool/local_server/main.py's /execute endpoint. The endpoint accepts JSON with attacker-controlled `command` and `shell` parameters, passes them directly to subprocess.run(shell=shell) without validation or authentication, and returns command output in the HTTP response. With shell=true, full shell syntax including chaining (&&, |, etc.) is supported. Network attack vector; no authentication required.

## Affected products

- HKUDS AnyTool 0.1.0

## Timeline

- 2026-09-30: disclosed: Vulnerability published on NVD
- 2026-02-28: other: Affected commit 506430fec in main branch

## References

- [https://github.com/HKUDS/AnyTool/](https://github.com/HKUDS/AnyTool/)
- [https://github.com/HKUDS/AnyTool/issues/6](https://github.com/HKUDS/AnyTool/issues/6)