The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the codex exec CLI tool to perform image generation tasks. It implements security controls by requiring the --sandbox workspace-write parameter and explicitly forbidding the use of flags that would bypass security approvals or the sandbox environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided prompts to generate assets. It mitigates injection risks by using quoted shell heredocs (<<'CODEX_IMAGE_REQUEST') to ensure prompt content is not interpreted as shell commands by the host environment.
Ingestion points: Prompt and Change fields in SKILL.md template blocks.
Boundary markers: Quoted heredoc delimiters are used to isolate user-supplied instructions from the shell.
Capability inventory: Uses codex exec with a restricted workspace-write sandbox.
Sanitization: Instructions require resolving all paths to absolute paths before execution to prevent directory traversal or ambiguity.
[EXTERNAL_DOWNLOADS]: The skill provides reference links to source code and implementation details hosted on the official GitHub repository of a trusted organization.

codex-image-generation