The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install and execute the k8spider tool directly from a non-trusted GitHub repository using go install github.com/Esonhugh/k8spider@latest. This pattern fetches and executes arbitrary code from an unverified external source at runtime.
[EXTERNAL_DOWNLOADS]: The skill requires downloading several external dependencies and tools from unverified sources, including scapy via pip, dsniff and ngrep via apt, and references the kube-dnsspoof project for DNS hijacking.
[DATA_EXFILTRATION]: The skill provides explicit instructions and Python scripts to perform ARP spoofing and MITM attacks to intercept sensitive data, specifically targeting API tokens, Bearer tokens, and database credentials from unencrypted internal traffic.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands for network manipulation and discovery, including nslookup, dig, arp, and tcpdump. It also includes commands to enable IP forwarding (echo 1 > /proc/sys/net/ipv4/ip_forward), which is a prerequisite for routing intercepted traffic through the attacker's container.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection.
Ingestion points: The agent processes output from DNS queries (nslookup, dig), system network files (/etc/resolv.conf), and HTTP response bodies from discovered services.
Boundary markers: No boundary markers or instructions to ignore embedded commands are present in the skill.
Capability inventory: The agent uses discovered service names and IPs to construct subsequent curl commands and tool invocations (k8spider, kubectl).
Sanitization: There is no evidence of sanitization or validation of the data discovered during the reconnaissance phase before it is used to drive further actions.

k8s-network-recon