cps-attack

Warn

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill contains hardcoded credentials for target systems in payloads.md. Specifically, it lists default and exploit-related passwords for Schneider Quantum PLCs (e.g., firefighter@somewhere, py8jdc@s5mb1, uucve@h6rbq) as part of its exploitation payloads.
  • [EXTERNAL_DOWNLOADS]: The skill frequently references and instructs the installation of external Python and Node.js packages for industrial protocol interaction and simulation.
  • Evidence includes pip install commands for conpot, minicps, cpppo, pyModbusTCP, snap7, pydnp3, S7CommPlusDriver, and BAC0 in guides/cps-attack-playbook.md and payloads.md.
  • Node.js package node-opcua is also recommended for installation.
  • [REMOTE_CODE_EXECUTION]: The guides/cps-attack-playbook.md file provides instructions to clone the OpenPLC project from a GitHub repository and execute its installation script (./install.sh), which involves running remote code locally to set up a simulation environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it is designed to ingest and process data from external, potentially adversarial Industrial Control System networks.
  • Ingestion points: Network traffic captures (PCAPs processed via tcpdump and zeek), Modbus register dumps, OPC UA endpoint data, and SQL historian database records.
  • Boundary markers: The provided scripts and methodologies lack explicit delimiters or instructions for the agent to ignore embedded commands within the ingested data.
  • Capability inventory: The skill utilizes high-privilege tools such as nmap, scapy, metasploit, and python3 for network manipulation and command execution.
  • Sanitization: There is no evidence of sanitization or validation of the data retrieved from external industrial devices before it is processed or used in decision-making logic.
  • [COMMAND_EXECUTION]: The skill extensively uses shell commands and Python scripts to interact with hardware and network protocols, including packet crafting with scapy and direct PLC manipulation using snap7 and cpppo.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 19, 2026, 04:36 PM