The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use sudo for administrative operations that modify host-level DNS configuration (e.g., sudo container system dns create).
[EXTERNAL_DOWNLOADS]: The skill involves downloading a binary installer package from Apple's official GitHub repository and fetching OCI images from external registries. It also describes functionality to download Linux kernels from user-provided remote URLs using the container system kernel set --tar command.
[DATA_EXFILTRATION]: The skill provides instructions for mounting host filesystem paths into containers and copying files between the host and container environments (container cp), which presents a risk of sensitive data access if host paths are not restricted.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from third-party OCI images and external registries.
Ingestion points: OCI image layers and registry metadata pulled from external sources (SKILL.md).
Boundary markers: The skill includes explicit safety warnings advising the user to "Inspect and trust the image source before pulling or running it" and "obtain explicit user approval" (SKILL.md).
Capability inventory: File read/write via container cp, network operations via registry interactions, and arbitrary command execution within containers via container exec and container run.
Sanitization: No specific content sanitization or isolation mechanisms for image content are described beyond the tool's VM-based isolation.
[NO_CODE]: The skill folder contains only markdown documentation and does not include any executable scripts or binary files.

apple-container