---
id: "CVE-2026-67543"
page: "https://infrastack.ai/live/CVE-2026-67543"
kind: "vulnerability"
aliases: ["CVE-2026-67543"]
title: "Heym: skill Python runs unisolated in backend, escalating to host root"
severity: "high"
cvss: 8.8
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: [{"name": "Heym", "fixed": [], "affected": "", "ecosystem": "vendor"}]
ecosystems: ["vendor"]
tags: ["ai-tooling", "docker", "heym", "python", "self-hosted"]
not_affected_if: ["no-docker-socket-mount", "trusted-skill-authors-only", "version-0.0.70-or-later"]
published_at: "2026-11-11T22:17:28+00:00"
first_seen_at: "2026-11-11T22:29:35+00:00"
updated_at: "2026-11-11T22:31:24+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-67543", "first_seen_at": "2026-11-11T22:29:35+00:00"}]
exploit_sources: []
---

# Heym: skill Python runs unisolated in backend, escalating to host root

Link for people: https://infrastack.ai/live/CVE-2026-67543 (this .md is for agents)

Heym's skill executor runs skill Python in the backend process with only a memory limit and a short environment denylist, instead of a Docker sandbox. A malicious or shared skill can reach the filesystem, the network, and non-denylisted secrets, and with the shipped Docker config (root backend plus mounted docker.sock) it can gain root on the host.

_Summary, check and fix written by a model from the sources; labelled as such._

## Affected

- vendor:Heym affected range not listed fixed no fixed version listed

## How to check

- Check the Heym version: versions below `0.0.70` are affected.
- Check whether the backend container runs as root and mounts `/var/run/docker.sock` (for example in docker-compose.yml).
- Check whether untrusted or shared users can create or run skills.

## Fix

- Heym: upgrade to `0.0.70` or later.

## Timeline

- 2026-11-11T22:29:35+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-67543

## Documents

- [advisory] Vendor advisory: github.com https://github.com/heymrun/heym/security/advisories/GHSA-hcv7-mg77-pg73
- [fix] Patch: github.com https://github.com/heymrun/heym/commit/2155057f4153a1e3a7b789d104f7fa96e92964e5
- [fix] Patch: github.com https://github.com/heymrun/heym/commit/28bb2901e4dea128ac493f67ee9478d230b7a3d7
- [fix] Release notes: github.com https://github.com/heymrun/heym/releases/tag/0.0.70

## From the sources

Heym builds and runs agentic systems and business-process automations on self-hosted infrastructure. Prior to version 0.0.70, heym hardens one untrusted-code path but not its sibling. The Python tool executor (python_tool_executor.py) treats user code as untrusted and mandates a Docker sandbox (no network, read-only fs, non-root, no docker socket), failing closed. The skill executor (skill_python_executor.py) runs skill Python via a local backend subprocess with no Docker isolation — only a 512 MB RLIMIT and a 7-key env denylist — so a skill executes inside the backend execution context, with filesystem access, network egress, and non-denylisted environment variables that should not be exposed to untrusted/shared skills. In the shipped Docker config (backend runs as root + /var/run/docker.sock mounted), this escalates to host root. This issue has been patched in version 0.0.70.

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_