skills/mukul975/anthropic-cybersecurity-skills/detecting-dnp3-protocol-anomalies/Gen Agent Trust Hub
detecting-dnp3-protocol-anomalies
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted network traffic data, which could contain malicious payloads designed to influence the agent.
- Ingestion points: The diagnostic script in
SKILL.mdreads PCAP files viascapy, whilescripts/agent.pyingests Zeek DNP3 logs. - Boundary markers: The skill does not implement delimiters or explicit instructions to treat the protocol data as untrusted content.
- Capability inventory: The scripts are restricted to parsing and reporting analysis to standard output; they do not possess capabilities for network exfiltration or system modification.
- Sanitization: Protocol fields and log entries are parsed and output to the agent's context without escaping or validation against injection patterns.
Audit Metadata