The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download the signed .pkg installer for the container tool from the official Apple GitHub releases page. It also documents capabilities for fetching images and Linux kernels from remote registries and URLs.
[COMMAND_EXECUTION]: The skill is designed to facilitate the execution of a wide range of container lifecycle and system management commands (e.g., run, exec, system start), which is its primary function.
[PRIVILEGE_ESCALATION]: Administrative privileges (sudo) are required and documented for specific local system configuration tasks, such as creating or deleting DNS resolver entries via container system dns.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface as the agent processes external, untrusted data from container logs, inspection metadata, and resource statistics. This risk is addressed by an explicit 'Safety Gate' instructing the agent to obtain user approval for any state-changing actions.
Ingestion points: container logs, container inspect, container stats, and container list (referenced in references/commands.md).
Boundary markers: The SKILL.md includes a 'Safety Gate' section requiring the agent to explain command impacts and obtain explicit user consent before execution.
Capability inventory: The skill can manage container lifecycles, background system services, virtual networks, and persistent volumes.
Sanitization: The documentation does not specify explicit output sanitization or filtering for container-generated data.
[DYNAMIC_EXECUTION]: The documented CLI includes a plugin mechanism that executes external binaries (container-<plugin>) when a non-standard command is invoked.
[NO_CODE]: The skill package is composed entirely of Markdown documentation and reference files; it does not contain any executable scripts (e.g., Python or Shell) in its distribution.

apple-container