powerpoint-reader

id	powerpoint
name	PowerPoint Reader
description	Analyze PowerPoint (.pptx) presentations by converting them to PDF via Keynote and reading the slides visually.
apps	["Keynote"]
tasks	["run_shell_command","read_file","spotlight_search"]
essential_tasks	[]
examples	["Summarize this PowerPoint presentation","What are the key points in slides.pptx?","Read the presentation on my Desktop and list the action items","Compare these two PowerPoint decks","Find the latest presentation I worked on and summarize it"]
safe_defaults	{}
confirm_before_write	[]
requires_permissions	["Keynote (Automation)"]

Behavior Notes

How It Works

macOS cannot read .pptx files as text — they are binary archives. To analyze a PowerPoint presentation, convert it to PDF using Keynote (pre-installed on macOS), then read the PDF visually. This preserves full layout, diagrams, charts, and screenshots.

Step 1: Find the File

If the user doesn't provide a path, use spotlight_search to locate it:

spotlight_search(file_name="presentation", content_type="presentation")

Step 2: Convert PPTX to PDF via Keynote

Use run_shell_command with this AppleScript. Replace the input/output paths:

osascript -e '
set inputFile to POSIX file "/path/to/presentation.pptx"
set outputFile to POSIX file "/tmp/presentation.pdf"
tell application "Keynote"
    open inputFile
    delay 5
    set theDoc to front document
    export theDoc to outputFile as PDF with properties {PDF image quality:Best}
    close theDoc saving no
end tell
'

• The delay 5 gives Keynote time to render the file. Increase to delay 10 for very large presentations.
• Keynote opens briefly during conversion — this is expected.
• Always write the PDF to /tmp/ to avoid cluttering the user's filesystem.
• Use a unique filename if processing multiple files (e.g., /tmp/deck_sales_q4.pdf).

Step 3: Read the PDF

Use read_file to read the converted PDF:

read_file(file_path="/tmp/presentation.pdf")

• Maximum 20 pages can be read at once.
• For larger decks, read in batches using the offset parameter and tell the user which slides you're covering.

Important Rules

• Never skip the conversion step. Reading a .pptx directly with read_file will produce garbage — always convert to PDF first.
• Clean up after yourself. Delete the temporary PDF from /tmp/ when done: run_shell_command(command="rm /tmp/presentation.pdf").
• Keynote opens .pptx natively. No plugins or extra software needed.
• Paths with spaces must be quoted. The AppleScript POSIX file handles this, but ensure the path string itself is correct.
• If Keynote fails, it may be because the file is corrupted or uses unsupported features. Tell the user and suggest opening it manually.

Export as Individual Slide Images (Alternative)

For very image-heavy decks or when you need per-slide analysis:

osascript -e '
set inputFile to POSIX file "/path/to/presentation.pptx"
set outputFolder to POSIX file "/tmp/slides"
tell application "Keynote"
    open inputFile
    delay 5
    set theDoc to front document
    export theDoc to outputFolder as slide images with properties {image format:PNG, skipped slides:false}
    close theDoc saving no
end tell
'

This creates slide001.png, slide002.png, etc. in /tmp/slides/. Read individual images with read_file.

Common Patterns

• "Summarize this deck" -> Convert to PDF, read all pages, provide summary with key points per slide.
• "What's on slide 5?" -> Convert to PDF, read with offset to target that page.
• "Find my latest presentation" -> spotlight_search(recently_used=True, content_type="presentation", days=7), then convert and read.
• "Compare two decks" -> Convert both to PDF, read both, provide comparison.