The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to interact with the local host environment using tools like the terminal, browser, and git directly. It specifically directs the agent to execute CLI commands such as codex exec and local scripts like spawn-codex-worker as part of its task delegation workflow.
[PROMPT_INJECTION]: The content includes instructions to override the agent's default model selections and workflow paths, substituting them with specific 'Codex' models and local tool configurations. These act as operational overrides for the agent's task-handling logic.
[DATA_EXPOSURE]: The skill mentions capturing 'human corrections' and 'unexpected outcomes' to generate structural mechanisms. While this creates a surface where external feedback influences the generation of scripts or lint rules, it is presented as a high-level operational principle for system improvement.

principle-encode-lessons-in-structure