The Agent Skills Directory

[SAFE]: The skill functions as a set of instructional guidelines for task delegation and model selection (routing) within the agent's workflow.
[DATA_EXPOSURE]: The skill mentions that the Codex CLI honors the configuration file at ~/.codex/config.toml. This is a neutral reference to how an external tool operates and does not include instructions for the agent to access, display, or exfiltrate the contents of this configuration file.
[INDIRECT_PROMPT_INJECTION]: The routing logic depends on analyzing external inputs such as task descriptions and workflows. Although this creates a surface where malicious data could theoretically influence model selection, the skill lacks the necessary permissions or tools (e.g., network access or file writes) to facilitate a harmful outcome from such an injection.

models