The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute codex exec to perform autonomous coding tasks. It also utilizes git for worktree management and mktemp to handle temporary output files.
[EXTERNAL_DOWNLOADS]: Fetches the @openai/codex CLI package from OpenAI's official registry using npm or brew for installation.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The codex exec command in SKILL.md accepts a free-form string as a task prompt which may contain untrusted data if the parent agent is delegating content from external sources (e.g., PR comments, issue descriptions).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to encapsulate the subagent's task prompt.
Capability inventory: The skill explicitly configures the workspace-write sandbox and permits optional network_access, granting the subagent the ability to modify the local filesystem and make network calls.
Sanitization: There are no instructions for sanitizing or escaping the content before it is interpolated into the subagent's execution command.

codex-subagent