scada-ics-security
Warn
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: MEDIUMPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill's methodology and technical guides (e.g., in guides/purdue-model-attack-paths-guide.md and guides/ics-network-assessment.md) suggest using 'sudo' for network scanning and packet capture operations with nmap and tcpdump. While standard in industrial security auditing to facilitate raw socket access, the use of administrative privileges increases the potential impact of command execution.
- [INDIRECT_PROMPT_INJECTION]: The skill includes scripts that process output from tshark and tcpdump (found in guides/ics-incident-response-guide.md and guides/scada-network-monitoring-guide.md). 1. Ingestion points: Network traffic parsed by tshark and displayed to the agent. 2. Boundary markers: Absent. 3. Capability inventory: Bash, Write, Edit, WebSearch, and WebFetch in SKILL.md. 4. Sanitization: The scripts perform basic parsing (splitting by tabs) but lack robust filtering of malicious content, creating a vulnerability surface where crafted protocol packets could potentially influence agent behavior.
- [COMMAND_EXECUTION]: Multiple guides include Python and Shell scripts that execute system utilities such as tshark, tcpdump, and nmap via subprocess calls to perform network forensics and monitoring.
- [EXTERNAL_DOWNLOADS]: The skill fetches vulnerability information from the NIST NVD API (services.nvd.nist.gov) and references industry-standard tools like 'conpot', both of which are considered trusted or well-known services.
- [CREDENTIALS_UNSAFE]: The Siemens S7 attack guide (guides/modbus-s7comm-attack.md) contains a list of common default vendor passwords (e.g., 'SYSTEM', 'S7', '00000000') used for security auditing and brute-force testing of industrial controllers.
Audit Metadata