highexperimentalLinuxOCIT1059.004

AI Agent Spawning Shell Interpreter

Detects AI agent runtimes (Python, Node) spawning interactive shell interpreters — a strong indicator of agent goal hijacking, prompt injection leading to code execution, or unsafe tool invocation.

Updated Jan 10, 2025 · OCI AI Security Team

agentic-aishell-executionlinuxociowasp-asi01

Problem Statement

AI agents running on Linux (Python/Node runtimes) should not spawn interactive shells under normal operation. Shell spawning from agent processes indicates goal hijacking, prompt injection, or unsafe code execution. Early detection prevents lateral movement and further compromise.

Sample Logs

{"timestamp":"2025-01-10T14:22:01Z","computer_name":"oci-worker-01","user":"agent_svc","image":"/bin/bash","command_line":"bash -i","parent_image":"/usr/bin/python3","parent_command_line":"python3 /opt/agent/run.py"}

Required Fields

image

command_line

parent_image

parent_command_line

user

computer_name

False Positives

·Legitimate automation scripts that spawn shells for system administration
·Build systems or CI/CD pipelines using Python to orchestrate shell tasks

Tuning Guidance

Whitelist known CI/CD service accounts and approved automation pipelines. Focus on interactive sessions (TTY-attached shells) to reduce false positives from batch automation.