ExcalidrawAutomate full library for LLM training.md has grown to become large, almost 2MB in size, eating up a whooping 500k+ tokens. I use this with Google’s aistudio and find that generated scripts are typically first time right. I personally had not experienced any issues with the very large training file, but recognize that this is not the most token efficient approach. I created an alternative version of the training file using Anthropic’s skill standard.
You’ll find the Excalidraw-Automate Skill in this folder on GitHub: Skill folder
Here’s my video explaining my script generation workflow using AI Studio. Note I do not have an AI subscription. This approach is still 100% free. Since the video’s release Google released newer Gemini models. Even though Gemini 3.5 Flash release is available in aistudio, I found that 3.1 Pro returns better quality scripts, so for now I am sticking to Gemini 3.1 pro.
AI Scripting Superpowers: Create ANY Obsidian-Excalidraw Script with Gemini 2.5 - Easy & Free!
If you are using Claude Code with this skill, I’d love to hear your experience with the skill. Anything I need to tweak to make it better?