The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill makes extensive use of sudo to perform administrative tasks, which is a high-privilege operations surface.
Commands include sudo iptables to view and modify firewall chains (FW_WRAPPER).
Accesses kernel ring buffer logs using sudo dmesg to inspect blocked traffic.
Uses sudo cat to read protected log files located in /tmp/ directories.
Executes the main application with elevated privileges using sudo awf during debug workflows.
[COMMAND_EXECUTION]: The skill utilizes powerful system commands to interact with the host and containers.
Uses docker exec to run commands inside managed containers, including reading sensitive configuration files like /etc/squid/squid.conf and viewing environment variables via env.
Includes cleanup routines that perform destructive actions such as docker rm -f, docker network rm, and rm -rf /tmp/awf-*.
Uses nc (netcat) for network connectivity testing between containers.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a vulnerability surface where malicious network traffic could influence the agent's behavior.
Ingestion points: The skill reads and parses Squid proxy access logs (/var/log/squid/access.log) which contain domain names and headers from external network requests.
Boundary markers: None. The skill directly reads and displays log content without delimiters or instructions to ignore embedded content.
Capability inventory: The skill has the capability to execute shell commands via docker exec, modify firewall rules via sudo iptables, and delete files/containers.
Sanitization: None. The logs are processed using standard shell tools like awk, grep, and tail without validation or filtering of the log content before it is presented to the agent.

debug-firewall