The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for performing a container-to-host escape (MITRE ATT&CK T1611) by scheduling a privileged pod that mounts the host's root filesystem. This is presented as a proof-of-concept for use in authorized lab environments.
[DATA_EXFILTRATION]: The instructions and scripts demonstrate how to retrieve sensitive Kubernetes secrets and service-account tokens, which are high-value credentials within a cluster environment.
[EXTERNAL_DOWNLOADS]: Fetches the rbac-police auditing tool from Palo Alto Networks' official GitHub releases.
[EXTERNAL_DOWNLOADS]: Uses the kubectl krew plugin manager to install security utilities such as who-can (Aqua Security) and rbac-lookup (Fairwinds) from established third-party sources.
[REMOTE_CODE_EXECUTION]: Executes the downloaded rbac-police binary after modifying its permissions with chmod +x to automate the discovery of escalation paths.
[COMMAND_EXECUTION]: The scripts/agent.py script utilizes the subprocess.run function to programmatically execute kubectl commands for cluster resource enumeration.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external cluster data (RBAC configurations) without boundary markers or sanitization.
Ingestion points: scripts/agent.py ingests data from kubectl get commands.
Boundary markers: None identified in the instructions or scripts.
Capability inventory: The script has the ability to execute subprocesses through kubectl (defined in scripts/agent.py).
Sanitization: No filtering or validation of cluster-provided data is performed before processing.

auditing-kubernetes-rbac-privilege-escalation