The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Python's subprocess module in render_codex.py, doctor.py, and settings.py to interact with system utilities. Specifically, it calls the codex CLI to execute image generation tasks and the sips tool on macOS for image format conversion. These calls are restricted to expected CLI tools and are functional requirements for the skill's operation.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as its primary workflow involves reading and summarizing user-provided files or articles to generate comic scripts. This is a standard risk for content generation skills.
Ingestion points: SKILL.md describes a workflow where the agent reads user-provided articles or documents to extract core concepts (Step 1).
Boundary markers: The workflow does not explicitly define delimiters for external content during the initial summarization phase.
Capability inventory: The skill possesses the capability to execute commands via the codex CLI (render_codex.py) and write to the local file system (settings.py, _data.py).
Sanitization: The instructions focus on content extraction and formatting; no specific sanitization or escaping of the ingested text is documented before it is interpolated into internal prompts.
[EXTERNAL_DOWNLOADS]: The documentation and setup scripts reference external dependencies, including the Codex CLI and the Python Pillow library. These are well-known developer tools and libraries required for the skill to perform image processing and generation.

comic-maker