The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted network data from RouterOS using tools like tshark and Wireshark. This creates an attack surface where malicious network packets could contain instructions intended to influence the agent's behavior during protocol analysis.
Ingestion points: Network interfaces monitored via tshark -i any and /tool/sniffer packet streams.
Boundary markers: No specific delimiters or safety instructions are provided to sanitize or isolate the packet content from the agent's prompt context.
Capability inventory: The skill utilizes curl for REST API interactions, tshark for live capture, and qemu-system-x86_64 for environment emulation.
Sanitization: No packet sanitization or filtering logic is suggested beyond protocol-level filters.
[PRIVILEGE_ESCALATION]: The host-side setup instructions require administrative privileges to install tools and modify security policies.
Evidence: The file references/tzsp-receivers.md includes commands for sudo apt install tshark, sudo ufw allow 37008/udp, and sudo iptables -A INPUT -p udp --dport 37008 -j ACCEPT.
[COMMAND_EXECUTION]: The skill provides numerous CLI command patterns for the agent to execute on both the host and the router.
Evidence: Commands for qemu-system-x86_64 VM management, curl for remote administration, and tshark/tcpdump for network monitoring are provided across both files.

routeros-sniffer