The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines the atmos container exec command, which allows the agent to run arbitrary shell commands inside container components. This provides significant control over the container environment and could be misused if not properly audited.
[INDIRECT_PROMPT_INJECTION]: The skill processes external configuration data which could contain malicious instructions designed to influence the agent.
Ingestion points: Atmos stack manifests, Dockerfiles, and CI Markdown templates defined via the ci.templates.container.image configuration mentioned in SKILL.md.
Boundary markers: The documentation does not describe the use of specific delimiters or explicit warnings to prevent the agent from following instructions potentially embedded within these configuration files.
Capability inventory: The skill uses the atmos CLI to perform filesystem operations for build contexts, network requests for registry push/pull operations, and direct container command execution via the exec subcommand.
Sanitization: There is no indication that the content of the manifests or templates is sanitized or validated against a strict schema before being interpreted by the agent.

atmos-container