-
Notifications
You must be signed in to change notification settings - Fork 0
Description
Problem
pptx show needs to present slide contents in a way that is useful for an agent across different tasks:
- Editing a data label: need to find the tiny "954" text box
- Rewriting narrative: need the bullet-point text shapes
- Understanding content: need a text summary, not positions
- Replicating layout: need everything including decorative shapes
- Updating placeholders: need to know which shapes are structural
Current flat list of 78 shapes (on a complex slide) does not serve any of these well.
Design questions
-
Filtering vs. showing everything: Filtering by word count wrongly hides chart data labels. But 78 shapes in a flat list is overwhelming. What is the right default?
-
Sort order:
(top, left)reading order seems right, but does not capture visual grouping (e.g. shapes forming a "row" of a waterfall chart). -
Modes vs. one output: Should
showhave modes (--text,--all,--structure) or one good default with drill-down viashow <file> <slide> <shape>? -
Placeholder constraints: Placeholders (title, subtitle, tracker) have implicit rules:
- Title: text only, 1-2 lines, do not change font/size/position, one per slide
- Subtitle: text only, 1 line, same constraints
- Content placeholders: flexible — use as-is or delete for custom content
- Should these rules be enforced in the tool or documented in SKILL.md?
-
Numbering: Placeholders use
ph0,ph1(stable IDs). Content shapes need stable references too — index is random, name is often duplicated ("Content Placeholder 4" appears 20x). Position-based numbering?
Reference
Complex slide example (slide 5 of Tiber deck): 78 shapes total
- 4 placeholders (title, subtitle, tracker, body)
- 1 chart (waterfall)
- 13 text shapes (≥3 words)
- 25 labels (1-2 words — chart axis labels, data values)
- 33 decorative (connectors, rectangles, freeforms)
- 3 groups
Next steps
Pick a design, prototype it, test on real decks.