The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the exec tool with PTY support to invoke the codex CLI, which is designed to modify the local filesystem and run shell commands.
[REMOTE_CODE_EXECUTION]: The --yolo execution flag provides a mode that explicitly disables the security sandbox and removes manual approval requirements for code execution and file changes, allowing an external agent to perform unverified operations.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @openai/codex global NPM package as a prerequisite for its functionality.
[COMMAND_EXECUTION]: Encourages the use of background processes (background:true) for long-running tasks, which permits the agent to execute system-level operations without continuous user oversight.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from project repositories.
Ingestion points: Source code files, documentation, and pull requests within the specified workdir are read and processed by the agent.
Boundary markers: There are no boundary markers or instructions to the agent to ignore embedded commands within the processed files.
Capability inventory: The agent has capabilities for filesystem modification and shell command execution via the codex CLI.
Sanitization: No sanitization or input validation is performed on the files ingested from the project directory before they are processed by the LLM.

codex-agents