# CVE-2026-55376: Heym Python tool sandbox escape in Agent workflows

Source: https://junglewise.ai/threats/cve-2026-55376-heym-builds-and-runs-agentic-systems-and-business-process  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-13. Updated: 2026-10-16  
License: CC BY 4.0 with a link back

CVE-2026-55376 · Severity: high · CVSS 8.8 · Published 2026-10-13

Heym is a platform for building and running automated business workflows on self-hosted infrastructure. An authenticated user with permission to create or edit Agent workflows can bypass the Python code sandbox and execute arbitrary operating-system commands on the backend server. This allows an attacker to read files, modify data, or compromise the entire system running Heym.

Technologies: [Heym](https://junglewise.ai/threats/technologies/heym-heym). Vendors: [Heym](https://junglewise.ai/threats/vendors/heym).

## Quick answers

**Is CVE-2026-55376 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-55376 being exploited in the wild.

**How severe is CVE-2026-55376?** High severity, with a CVSS score of 8.8, published 13 October 2026.

**Which products and versions are affected?** Heym prior to 0.0.40.

**Is there a fix for CVE-2026-55376?** Yes: fixed in 0.0.40. Until then: Restrict workflow creation and editing permissions to trusted users only until an upgrade to version 0.0.40 is deployed. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-55376 attacked?** It is attacked over the network.

**Am I affected by CVE-2026-55376?** Check whether you run Heym 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.

## How to fix

- Fixed in 0.0.40

Restrict workflow creation and editing permissions to trusted users only until an upgrade to version 0.0.40 is deployed.

## Technical details

An authenticated user can create a custom Python tool in an Agent workflow that circumvents the blacklist-based sandbox in backend/app/services/python_tool_executor.py by accessing operating-system and process capabilities through public attributes of allowed standard-library modules (e.g., glob.os.system). The vulnerable code uses exec() to run the supplied tool without proper isolation, allowing OS commands and filesystem operations to run as the backend application user. The fix isolates Python tool execution in a hardened Docker container with an import allowlist, attribute filtering, and network isolation.

## Affected products

- Heym Heym prior to 0.0.40

## Timeline

- 2026-10-13: disclosed
- 2026-06-12: patched

## References

- [https://github.com/heymrun/heym/commit/b36d795c6ea841b8f1b511a0952e274487ee4ef3](https://github.com/heymrun/heym/commit/b36d795c6ea841b8f1b511a0952e274487ee4ef3)
- [https://github.com/heymrun/heym/releases/tag/v0.0.40](https://github.com/heymrun/heym/releases/tag/v0.0.40)

## Related threats

- [Heym command injection via MCP stdio transport](https://junglewise.ai/threats/cve-2026-67546-heym-builds-and-runs-agentic-systems-and-business-process)
- [Heym HTTP node server-side request forgery](https://junglewise.ai/threats/cve-2026-67545-heym-builds-and-runs-agentic-systems-and-business-process)
- [Heym workflow engine arbitrary code execution via Playwright node](https://junglewise.ai/threats/cve-2026-67544-heym-builds-and-runs-agentic-systems-and-business-process)
- [Heym Python skill execution privilege escalation](https://junglewise.ai/threats/cve-2026-67543-heym-builds-and-runs-agentic-systems-and-business-process)
- [Heym server-side request forgery in MCP transports](https://junglewise.ai/threats/cve-2026-67542-heym-builds-and-runs-agentic-systems-and-business-process)