The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the @openai/codex package via the npm package manager. This package is maintained by OpenAI, a well-known service, and its installation is a prerequisite for the skill's functionality.
[COMMAND_EXECUTION]: The skill uses the codex CLI to perform automated tasks. It specifically uses the --sandbox workspace-write flag, which is a recommended security practice to restrict the subagent's write access to the designated project workspace.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by design, as it passes user-supplied instructions directly to the Codex CLI. While this is necessary for its operation, it creates a potential vulnerability if the input contains malicious instructions intended to influence the subagent's actions.
Ingestion points: Task prompts provided to the codex exec command in SKILL.md.
Boundary markers: No explicit delimiters or boundary markers are suggested to encapsulate the user-provided prompt content.
Capability inventory: The subagent has the capability to modify files and perform coding tasks within a sandboxed environment.
Sanitization: The skill does not implement or suggest any sanitization of the input prompt before it is passed to the subagent.

codex-subagent