Suspicious OpenClaw AI Agent Spawning Command Shell via Process Creation

PremiumReviewedSigma · Medium · v1
Category
process_creation
Author
HuntRule
Published
2026-05-19
Updated
2026-08-28

ATT&CK techniques

Execution
  1. Recon

  2. Resource Dev

  3. Initial Access

  4. Persistence

  5. Priv Esc

  6. Defense Evasion

  7. Cred Access

  8. Discovery

  9. Lateral Movement

  10. Collection

  11. C2

  12. Exfiltration

  13. Impact

What it detects

This rule detects the OpenClaw AI agent binaries openclaw, clawdbot, or moltbot spawning a Unix command shell, the execution pattern that follows a poisoned agent skill being loaded. Adversaries plant malicious skills so the agent runs attacker commands with the user privileges, so an AI agent process launching sh, bash, or zsh is a strong sign of skill-based code execution.

Related detections9 linkedT1059.004 — drag to rearrange
Malicious Shell Payload Piped from curl to zsh
Possible Bitbucket Pre-Auth RCE via git archive exec Null-Byte Injection (CVE-2022-36804) (via webserver)
Malicious TeamTNT Docker Gatling Gun Initialization Script (via process_creation)
Suspicious Shell Spawned by PostgreSQL Server Process (via process_creation)
Malicious Remote Script Piped Directly to a Shell (via process_creation)
HamsaUpdate Linux Payload Download via Wget Piped to Bash (via process_creation)
Suspicious Remote Script Execution via Wget or Curl Piped to Shell (via process_creation)
Malicious Bash Dev-TCP Reverse Shell via Shell (via process_creation)
Obfuscated Base64-Decoded Payload Piped to a Shell (via process_creation)
Suspicious OpenClaw AI Agent Spawning Command Shell via Process Creation
Pivot detection · T1059.004 · 9 related

Detection content is published as a reviewed draft. Tune thresholds and exclusions against your own telemetry before enabling this rule for alerting.