The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py utilizes subprocess.run(shell=True) with string interpolation in functions such as run_tshark, capture_live, and apply_display_filter. This pattern is susceptible to shell command injection if inputs like pcap_path, interface, or display_filter are derived from untrusted sources without validation.\n- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection. It extracts and processes data from network packets, such as HTTP User-Agents, DNS queries, and TLS SNI fields, which are then returned as text to the AI agent. Maliciously crafted network traffic could contain instructions designed to influence the agent's logic during analysis.\n
Ingestion points: Network packet data from PCAP files enters the agent context via extract_http_requests, extract_dns_queries, and extract_tls_info in scripts/agent.py.\n
Boundary markers: None present; data is extracted and printed as raw strings to stdout without delimiters.\n
Capability inventory: The skill can execute arbitrary tshark commands, perform live captures, and export objects to the local filesystem via scripts/agent.py.\n
Sanitization: No sanitization, escaping, or validation is performed on the strings extracted from packet fields before they are returned to the agent.

analyzing-network-traffic-with-wireshark