The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses several internal Python scripts (generate_ppt.py, edit_slide.py, serve_viewer.py, package_pptx.py) to manage the presentation lifecycle. These scripts perform standard file system operations and process execution within the local workspace.- [EXTERNAL_DOWNLOADS]: The skill relies on standard third-party libraries such as google-genai for model interaction, python-pptx for document creation, and Pillow for image processing. These are common and reputable dependencies for this application type.- [PROMPT_INJECTION]: The workflow incorporates user feedback from a local review server into image generation prompts via edit_slide.py. While this is an intended feature for iterative editing, it represents an indirect prompt injection surface where user-supplied text influences model behavior.
Ingestion points: User feedback is captured by serve_viewer.py and stored in slides_plan.json.
Boundary markers: Absent; feedback is passed directly as an instruction to the image generation model.
Capability inventory: The skill can execute local scripts, write files via write_file/edit_file tools, and perform network requests to the Gemini API.
Sanitization: No explicit sanitization or validation is applied to user feedback before it is interpolated into model prompts.

nano-banana