The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes structured prompting guidelines and a 'specificity policy' to ensure image quality. These are functional instructions aimed at improving model output and do not contain attempts to override or bypass AI safety constraints. Additionally, while the batch mode processes external prompts from files, this is a standard operational feature for the described automation use case.
[CREDENTIALS_UNSAFE]: While the skill mentions the OPENAI_API_KEY as a prerequisite for using the CLI tool fallback, no actual credentials or API keys are hardcoded in the file. It also explicitly highlights that the primary built-in tool operates without the need for an API key.
[EXTERNAL_DOWNLOADS]: The documentation suggests installing standard, well-known Python libraries (openai, pillow) via uv pip to support the CLI scripts. This is a common requirement for the described functionality and involves reputable packages from public registries.
[COMMAND_EXECUTION]: The skill provides instructions for the agent and the user to execute shell commands such as python for script activation and cp (copy) for asset management. These operations are routine for local development environments and are scoped to the project's designated worktree and data directories.

codex-imagegen