The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill provides commands utilizing sudo to modify system-wide configuration files in /etc/sysctl.d/ to adjust user namespace limits and unprivileged network port thresholds. It also suggests using sudo usermod to modify subordinate UID and GID mappings on the host.
[EXTERNAL_DOWNLOADS]: The instructions include the installation of the podman-compose package using pip install, which fetches and installs third-party code from a remote registry.
[PERSISTENCE]: The skill guides the user to establish persistence for containerized services using systemctl --user enable and by enabling user lingering with loginctl enable-linger $USER, allowing processes to persist after a user logs out.
[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection by processing external configuration files that may contain instructions.
Ingestion points: Processes local file content during image builds (podman build) and YAML deployment files (podman kube play).
Boundary markers: There are no specific instructions or delimiters provided to prevent the agent from following directives embedded within the Dockerfiles or Kubernetes manifests.
Capability inventory: The skill possesses capabilities for file system interaction, network communication (image registry operations), and containerized code execution.
Sanitization: No sanitization or validation mechanisms are described to verify the integrity of the content within the processed configuration files.

podman