Chart To Drive

Skill: Chart-to-Drive Uploader

Purpose

Standardized workflow for uploading local chart PNGs to Google Drive and making them available for insertion into Google Docs and Slides. One standard flow for upload, permissions, and URL construction, with no public file hosts involved.

When to Apply

Automatically whenever:

• Chart PNGs need to be inserted into Google Docs or Google Slides
• The google-doc-creator or google-slides-creator agent needs chart URLs
• User asks to "upload charts" or "add images to the doc/slides"
• User encounters "permission denied" or "image URL not accessible" errors when building Google Workspace documents — this usually means charts weren't uploaded to Drive first

When NOT to use this skill:

• For single-image uploads where you just need one Drive URL (use mcp__google-docs__upload_image_to_drive directly)
• When building Google Docs via the python-docx → upload workflow (that embeds images in the .docx file, no Drive URLs needed)

Workflow

Step 1: Collect chart files

Identify all chart PNGs that need uploading. Standard location: outputs/charts/.

import os
chart_dir = "outputs/charts"
charts = [(f, os.path.join(chart_dir, f)) for f in sorted(os.listdir(chart_dir)) if f.endswith('.png')]

Step 2: Upload each chart to Drive directly

Use the Google Docs MCP upload_image_to_drive tool with the LOCAL file path. It uploads straight to the user's Drive and sets just enough link access for the Docs/Slides APIs to fetch the image. No public file host is involved at any point; never route charts through one (the analysis may contain confidential numbers).

for each (filename, filepath) in charts:
    mcp__google-docs__upload_image_to_drive(image_path=filepath)
    # Returns the Drive file ID; record it in the chart map

If that tool is unavailable, upload with the Drive API from Python using local credentials (files().create with MediaFileUpload), then set reader link access.

Optional: Create a "{dataset_name} - Charts" folder first and pass its ID as the parent so uploads stay organized.

Step 3: Build the URL map

Return a mapping of chart filename to both URL formats:

chart_map = {
    "01_height_crossover.png": {
        "drive_id": "1abc...",
        "drive_url": "https://drive.google.com/uc?id=1abc...&export=download"
    },
    ...
}

For Google Docs: Use drive_url (permanent, works with insert_doc_image) For Google Slides: Use drive_url (permanent, works with createImage)

URL Format Reference

Rules

1. Drive only, always. Never upload charts to public file hosts (tmpfiles, imgur, postimages, or similar); the user's Drive is the only image host.

2. Set public permissions immediately. Google Docs/Slides API cannot access private Drive files. The MCP upload_image_to_drive tool automatically sets public access — no additional permission call needed.

3. Batch uploads in a single Python script. Don't make separate Bash calls per chart — one script handles all uploads efficiently.

4. Print the chart map. Always output the full filename → drive_id mapping so subsequent agents can reference charts by name.

5. Check for existing uploads. Before re-uploading, search Drive for files with the same name in the expected folder. If you find existing files with matching names uploaded in the last 24 hours, reuse those Drive IDs instead of creating duplicates. Only re-upload if the file is missing or very old.

Error Recovery

If the MCP upload tool fails or is missing:

• Upload with the Drive API from Python using local credentials (files().create + MediaFileUpload), then set reader link access.
• If no Drive access works at all: stop, keep the charts local, tell the user which files need manual upload. Never fall back to a public file host.

If a Drive upload returns "permission denied" or "authentication failed":

• Verify Google Workspace MCP authentication status
• Run mcp__google-docs__authorize_google_docs() to re-authenticate
• Retry the upload after successful auth