The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied configuration files which could contain embedded instructions aimed at influencing the agent's behavior.
Ingestion points: The agent accepts user-provided JSON content via the 'Validate existing config' and 'Modify existing config' workflows described in SKILL.md.
Boundary markers: The skill includes a specific 'Security (must follow)' section in SKILL.md that instructs the agent to treat user-provided content as untrusted data and not as execution instructions.
Capability inventory: The skill utilizes the mirrord CLI (via shell) and MCP server tools (validate_config, explain_config_option) to process the data.
Sanitization: No programmatic sanitization is applied; the skill relies on instruction-following to separate data from instructions.
[COMMAND_EXECUTION]: The skill utilizes the local mirrord CLI to execute the verify-config command on temporary files. This is a standard functional requirement for the skill's purpose.
[SAFE]: The skill refers to official documentation at mirrord.dev and references container images from ghcr.io/metalbear-co/mirrord. These resources are consistent with the identified vendor, MetalBear.

mirrord-config