Update PPTX skill to enable DOM-based manipulation to reduce token consumption?

## Update PPTX skill to enable DOM-based manipulation to reduce token consumption?

The DOM-based pattern used in the Claude in PowerPoint add-on could be more token-efficient than the current skill approach. The main difference is in the orientation phase : the current skill reads each slide’s full XML to identify what to edit, while the DOM approach returns only lightweight metadata indexed by id, so the LLM never needs to read content just to navigate.

Would it make sense to explore adopting this approach in the pptx skill? The question is also relevant for the other document related skills.