Free AI Safety Workshop — in collaboration with explainx.aiSave your seat
AgentbeamBeta
AI agent vulnerability · OpenClaw

CVE-2026-32063: OpenClaw Improperly Neutralizes Line Breaks in systemd Unit Generation Enables Local Command Execution (Linux)

highCVSS 4.0 8.6GHSA-vffc-f7r7-rx2w

Summary A command injection vulnerability exists in OpenClaw’s Linux systemd unit generation path. When rendering Environment= entries, attacker-controlled values are not rejected for CR/LF, and systemdEscapeArg() uses an incorrect whitespace-matching regex. This allows newline injection to break out of an Environment= line and inject standalone systemd directives (for example, ExecStartPre=).

Published Mar 3, 2026 · Updated Mar 16, 2026

Affected packages

openclaw

Ecosystem
npm
Affected
<= 2026.2.19-2
Fixed in
2026.2.21
Check
npm ls openclaw

Advisory details

From the GitHub Advisory Database (CC BY 4.0).

Summary

A command injection vulnerability exists in OpenClaw’s Linux systemd unit generation path. When rendering Environment= entries, attacker-controlled values are not rejected for CR/LF, and systemdEscapeArg() uses an incorrect whitespace-matching regex. This allows newline injection to break out of an Environment= line and inject standalone systemd directives (for example, ExecStartPre=). On service restart, the injected command is executed, resulting in local arbitrary command execution (local RCE) under the gateway service user.

---

Details

The issue is in src/daemon/systemd-unit.ts:

  • renderEnvLines(...) builds:
  • Environment=${systemdEscapeArg(${key}=${value})}
  • No CR/LF validation is enforced for environment keys/values before writing unit lines.
  • systemdEscapeArg(...) uses:
  • /[\\s"\\\\]/
  • In this regex, \\s is interpreted as a literal backslash + s, not a whitespace character class.

As a result, whitespace detection/quoting behavior is incorrect. Because systemd parses unit files line-by-line, a newline inside an environment value can inject an additional directive line. Example rendered output:

Environment=INJECT=ok
ExecStartPre=/bin/touch /tmp/oc15789_rce

At restart time, systemd executes ExecStartPre, enabling command execution.

Relevant code path/components involved in exploitation chain:

  • src/daemon/systemd-unit.ts
  • src/commands/daemon-install-helpers.ts
  • src/config/env-vars.ts
  • src/config/zod-schema.ts

Trigger conditions:

  1. Attacker can influence config.env.vars (directly or indirectly).
  2. Install/reinstall path is invoked to write/update the unit.
  3. Service restart occurs (systemctl --user restart ...).

---

PoC

Environment: Linux host with systemd user services enabled.

  1. Configure a malicious environment value in OpenClaw config (config.env.vars), including a newline and injected directive:
  2. Key: INJECT
  3. Value:
ok
ExecStartPre=/bin/touch /tmp/oc15789_rce
  1. Install/reinstall the gateway service (fixed port as requested):
openclaw gateway install --port 15789 --force
  1. Inspect the generated user unit file (default path):
~/.config/systemd/user/openclaw-gateway.service

Verify that an injected standalone line exists:

ExecStartPre=/bin/touch /tmp/oc15789_rce
  1. Reload and restart user service:
systemctl --user daemon-reload
systemctl --user restart openclaw-gateway.service
  1. Confirm command execution side effect:
ls -l /tmp/oc15789_rce

---

Impact

This is a local command execution vulnerability in OpenClaw’s systemd unit generation during install/reinstall flows.

  • Type: Command injection via newline/directive injection in unit file generation.
  • Execution context: Runs with the same privileges as the OpenClaw gateway service user.
  • Affected users: Linux deployments using systemd user services where an attacker can control config.env.vars and trigger install/reinstall.

Fix Commit(s)

  • 61f646c41fb43cd87ed48f9125b4718a30d38e84

Credited to tdjackey.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. No analysis has been written for this entry yet. Methodology

More OpenClaw vulnerabilities

All OpenClaw entries