The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Python script, scripts/assemble_images.py, used for local image processing tasks such as vertical stitching and contact sheet generation. The script relies on the Pillow library and operates strictly on local files as part of the skill's primary functionality.
[PROMPT_INJECTION]: The SKILL.md file defines a HARD-GATE mechanism that uses imperative language to control agent behavior, specifically requiring user confirmation before calling generation tools. These are functional constraints designed to prevent unintended token usage and ensure quality control rather than malicious overrides.
[SAFE]: The skill exhibits an attack surface for indirect prompt injection as it processes user-provided visual descriptions to generate images.
Ingestion points: User requests for batch images, platform-specific style extraction, and reference image metadata.
Boundary markers: The skill enforces a mandatory 'HARD-GATE' confirmation step after a full project manifest is presented but before any tools are invoked.
Capability inventory: Access to the native image_gen tool and local file system execution for image manipulation scripts.
Sanitization: Generation is guided by structured templates (templates/codex-image-request.md) and strict 'six-part' prompt engineering rules defined in the reference guides to maintain control over the generated content.

codex-image-service