The Agent Skills Directory

[SAFE]: The skill consists of instructional documentation and configuration without any executable scripts, binary files, or external dependencies.
[COMMAND_EXECUTION]: The skill instructs the orchestrating agent on how to invoke the codex CLI for coding tasks. It includes patterns for both interactive and non-interactive use, as well as instructions for managing the local workspace environment and temporary git repositories.
[PROMPT_INJECTION]: The skill defines a workflow for processing external, untrusted content such as repository code, commit logs, and branch diffs, which creates a surface for indirect prompt injection.
Ingestion points: Reads files and git metadata from the current repository via codex review and codex exec commands.
Boundary markers: The instructions suggest using quoted task descriptions for commands to separate user input from shell arguments.
Capability inventory: The skill allows the agent to modify files in the workspace and execute Git commands through the codex utility.
Sanitization: No explicit sanitization or filtering is described in the provided instructions; it relies on the internal safety mechanisms of the Codex CLI tool.

codex-coding-agent