The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill enables high-privilege operations including process termination (kill -9, fuser -k) and network interface reconfiguration (ip addr, ip link). If triggered by untrusted user input, these can lead to system instability or Denial of Service.
[DATA_EXFILTRATION] (MEDIUM): Provides packet capture capabilities via tcpdump. Using flags like -A or -X allows the agent to read and potentially exfiltrate sensitive plaintext data (credentials, cookies) from the network stream.
[COMMAND_EXECUTION] (MEDIUM): Support for curl and wget allows arbitrary outbound network requests. This presents a vulnerability surface for Server-Side Request Forgery (SSRF) where an attacker could probe internal networks or exfiltrate data via HTTP.
[CREDENTIALS_UNSAFE] (LOW): Code examples include passing authentication tokens and passwords directly in command-line arguments (curl -u, Authorization: Bearer), which is a security anti-pattern as it risks exposing secrets in system logs or shell history.

network-tools