Google Docs

Reading Google Docs

Platform Compatibility

Use the native choice/question UI when the host provides one. If not available, ask the same question directly in chat with the listed options. When the instructions mention a Read tool, use the host's normal file/image reading capability.

Step 1: Ask User What They Need

ALWAYS ask the user first:

Ask with options:
- "Scan & summarize" - Get overview of entire document
- "Extract specific section" - Find and read a section WITH its images
- "Full document" - Load complete document content
- "Extract all images" - Download and view all images from document

If user selects "Extract specific section", ask follow-up: "Which section do you need?"

Step 2: Download Document

ALWAYS download BOTH formats for complete analysis:

# Extract DOC_ID from URL: https://docs.google.com/document/d/{DOC_ID}/edit?...

# Download markdown (for text)
curl -sL "https://docs.google.com/document/d/{DOC_ID}/export?format=md" -o /tmp/google-doc.md

# Download HTML (for images) - ALWAYS do this too
curl -sL "https://docs.google.com/document/d/{DOC_ID}/export?format=html" -o /tmp/google-doc.html

Step 3: Process Based on User Choice

Scan & summarize

1. Read first ~200 lines of markdown
2. Identify all headings
3. Count images (grep -c 'image' /tmp/google-doc.md)
4. Provide summary

Extract specific section (WITH images)

1. Find section in markdown: grep -n "section name" /tmp/google-doc.md
2. Read section content with the available file reader
3. Check for image references in that section: ![][imageN]
4. Extract those specific images from HTML and analyze them
5. Provide text + image analysis together

Example workflow:

# Find section line numbers
START=$(grep -n "## 4) Košík" /tmp/google-doc.md | cut -d: -f1)

# Read section and find image references
sed -n "${START},+100p" /tmp/google-doc.md | grep -o 'image[0-9]*'

# If images found (e.g., image5, image6), extract them:
# Images in HTML are in order, so image5 = 5th image
grep -o 'src="data:image/png;base64,[^"]*"' /tmp/google-doc.html | sed -n '5p' | sed 's/src="data:image\/png;base64,//' | sed 's/"$//' | base64 -d > /tmp/section-image1.png

Full document

1. Read entire markdown with pagination
2. Extract all images
3. Analyze images that appear important (screenshots, diagrams, mockups)

Extract all images

# Extract all images from HTML
grep -o 'src="data:image/png;base64,[^"]*"' /tmp/google-doc.html | nl | while read num line; do
  echo "$line" | sed 's/src="data:image\/png;base64,//' | sed 's/"$//' | base64 -d > /tmp/image$num.png
done

# Count extracted images
ls /tmp/image*.png 2>/dev/null | wc -l

View images with the available file/image reader:

Read(file_path: "/tmp/image1.png")

Important: Image References

In markdown export, images appear as ![][imageN] where N is the image number. These correspond to the Nth image in the HTML export.

When extracting a section:

• Look for ![][imageN] references in the text
• Extract those specific images by number
• Include image analysis in your response

Example

User: "Dej mi info o sekci Objednávky"

1. Download both md and html
2. Find section: grep -n "Objednávky" /tmp/google-doc.md
3. Read section content
4. Find image refs: ![][image2], ![][image3]
5. Extract images 2 and 3 from HTML
6. View and analyze the images
7. Provide complete response with text AND image descriptions

Notes

• Works only with publicly shared documents
• ALWAYS download both md and html - images are only in HTML
• Image numbers in markdown correspond to order in HTML
• The host model can usually view and analyze PNG/JPG images when given the image file path.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: antstudiocz
• Source: antstudiocz/ant-marketplace
• License: MIT
• Homepage: https://antstudio.cz

Install and usage instructions live in the source repository linked above.