The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes mirrord-up.yaml configuration files and CLI inputs, which are considered untrusted data that could contain malicious instructions.
Ingestion points: User-provided configuration files (mirrord-up.yaml) and CLI arguments are read into the agent context.
Boundary markers: The 'Security Boundaries' section in SKILL.md explicitly instructs the agent to treat these inputs as data, not instructions, and to avoid executing arbitrary logic from them.
Capability inventory: The skill has the capability to execute shell commands (mirrord up) and write to the local filesystem (mirrord up init).
Sanitization: The instructions mandate validating Kubernetes resource names against a specific regex (^[a-z0-9]([a-z0-9-]{0,61}[a-z0-9])?$) and rejecting shell metacharacters before any interpolation into shell commands.
[COMMAND_EXECUTION]: The skill's primary purpose is to guide the user in executing mirrord CLI commands.
Evidence: SKILL.md and README.md provide detailed instructions and example commands for mirrord up, mirrord up init, and various configuration flags.
Mitigation: The skill contains explicit instructions to warn users about the impact of 'replace' mode on shared clusters (scaling workloads to zero) and mandates presenting commands for review when not in autonomous mode.

mirrord-up