The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to invoke the codex CLI via bash. It documents high-privilege sandbox modes such as danger-full-access and the --dangerously-bypass-approvals-and-sandbox flag, which allows the sub-agent to perform actions without user confirmation or environment restrictions.
[EXTERNAL_DOWNLOADS]: The skill references the installation of the @openai/codex package from the npm registry. This is an official package from a well-known service.
[PROMPT_INJECTION]: The skill facilitates the processing of arbitrary user instructions by interpolating them into shell commands. This creates an attack surface for indirect prompt injection.
Ingestion points: The <task> placeholder and standard input in codex exec commands are populated by user-controlled text.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the command templates.
Capability inventory: The skill utilizes the Bash tool for command execution, allowing for file system access and system modification depending on the chosen sandbox mode.
Sanitization: No sanitization or validation of the user-provided prompt is performed before it is passed to the execution environment.

codex-exec