The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires several standard and well-known Python packages including 'pillow', 'google-genai', 'python-pptx', and 'python-dotenv' to handle image processing, API interaction, and presentation packaging. These are installed into the local environment via pip as part of the setup process.
[COMMAND_EXECUTION]: The workflow relies on executing local Python scripts ('generate_ppt.py', 'serve_viewer.py', 'edit_slide.py', 'package_pptx.py') to perform various stages of the presentation creation process. This includes launching a local HTTP server for interactive review.
[INDIRECT_PROMPT_INJECTION]: The skill implements an interactive review cycle where user feedback is collected via a local server and stored in 'slides_plan.json'. This feedback is subsequently used as a direct instruction for the Gemini image editing API in 'edit_slide.py'.
Ingestion points: User feedback is captured by 'serve_viewer.py' and stored in 'slides_plan.json' (located in the workspace root).
Boundary markers: Absent. The feedback text is interpolated directly into the instructions for the Gemini API.
Capability inventory: The 'execute' tool runs scripts that perform file reads/writes and call external APIs (Google Gemini).
Sanitization: Absent. User feedback is passed as-is to the image generation model.

nano-banana