The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/codex_imagegen.py executes the codex CLI tool using subprocess.run. This is the primary mechanism for the skill's functionality. The command is constructed using a list of arguments and runs with shell=False (default), which prevents OS-level shell injection.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it interpolates untrusted user input directly into the prompt context for the Codex agent.
Ingestion points: Untrusted data enters via the --prompt argument in scripts/codex_imagegen.py.
Boundary markers: The prompt template in build_generate_prompt and build_edit_prompt provides some structure but lacks explicit delimiters or warnings for the agent to ignore instructions embedded within the user's prompt.
Capability inventory: The skill possesses file system access (shutil.copy2) and the ability to execute the codex CLI tool (subprocess.run), which is a general-purpose agent executor.
Sanitization: No sanitization or escaping is performed on the user-provided prompt content before interpolation.

codex-imagegen