Google Drive Integration Skill

Composio-powered Google Drive skill. All OAuth is handled by Composio — you never see or store Google tokens, and the skill works the same whether the connected account is a personal Gmail or a Workspace account.

Setup (one-time)

1. Get a Composio API key — sign up at https://app.composio.dev/ and copy your API key.

2. Connect Google Drive in Composio — from the Composio dashboard, connect the Google Drive integration for a user_id of your choice (any string; e.g. your email or an internal tenant ID). Complete the Google OAuth flow.

3. Install the Python client — pip install composio (or let the platform install it via the requires.python metadata above).

4. Export env vars before running the skill:

   export COMPOSIO_API_KEY="ak_..."
   export COMPOSIO_USER_ID="the-user-id-you-used-in-step-2"
   

   If the skill runs on a platform that auto-injects Composio credentials, these are set for you — no manual export needed.

TENANT_ID is also accepted as a fallback for COMPOSIO_USER_ID (for platforms that already set that name).

Usage

Run the google_drive_api.py script via Bash for all Google Drive operations:

python skills/google-drive/google_drive_api.py <command> [options]

Commands

Discovery

# List files in a folder (defaults to My Drive root)
python skills/google-drive/google_drive_api.py list --folder-id FOLDER_ID --page-size 50 --order-by "modifiedTime desc"

# Search by name substring, MIME type, or a raw Drive query
python skills/google-drive/google_drive_api.py search --name-contains "weekly sales"
python skills/google-drive/google_drive_api.py search --mime-type "text/csv" --folder-id FOLDER_ID
python skills/google-drive/google_drive_api.py search --query "name contains 'depletions' and modifiedTime > '2026-01-01T00:00:00'"

# Find folders by name or parent
python skills/google-drive/google_drive_api.py find-folder --name "Retail Reports"
python skills/google-drive/google_drive_api.py find-folder --name-contains "2026" --parent-id FOLDER_ID

# List shared drives the account can see
python skills/google-drive/google_drive_api.py shared-drives

Add --all-drives to list and search when the files live in a shared (team) drive rather than My Drive.

Metadata

# File or folder metadata (name, mimeType, size, modifiedTime, parents)
python skills/google-drive/google_drive_api.py info --file-id FILE_ID

# Who has access to a file
python skills/google-drive/google_drive_api.py permissions --file-id FILE_ID

Reading File Content

# Download a binary/uploaded file (CSV, XLSX, PDF, ...) into the workspace
python skills/google-drive/google_drive_api.py download --file-id FILE_ID --save-to data/report.csv

# Export a Google-native file to a concrete format
python skills/google-drive/google_drive_api.py export --file-id FILE_ID --mime-type text/csv --save-to data/sheet.csv

--save-to writes the bytes to a local path in the workspace so you can parse them with Python (pandas.read_csv, openpyxl, ...). Without it, the command only prints the Composio file reference.

Export MIME types for Google-native files:

Writing & Organising

# Create a folder
python skills/google-drive/google_drive_api.py create-folder --name "Q3 Reports" --parent-id PARENT_ID

# Create a text file (CSV, TXT, MD) from a string or a local file
python skills/google-drive/google_drive_api.py create-file --name "summary.csv" --from-file out/summary.csv --parent-id FOLDER_ID --mime-type text/csv

# Rename / re-describe
python skills/google-drive/google_drive_api.py update-file --file-id FILE_ID --name "2026-07 depletions.csv"

# Move, copy, trash
python skills/google-drive/google_drive_api.py move --file-id FILE_ID --to-folder-id FOLDER_ID
python skills/google-drive/google_drive_api.py copy --file-id FILE_ID --name "backup.csv" --to-folder-id FOLDER_ID
python skills/google-drive/google_drive_api.py trash --file-id FILE_ID

# Share with a person or a domain
python skills/google-drive/google_drive_api.py share --file-id FILE_ID --type user --role reader --email person@example.com

Workflow Tips

• Folder first, then files: find-folder to resolve a folder name to an ID, then list --folder-id to see what is inside, then download/export the file you need.
• Extract IDs from URLs: folder IDs come after /folders/, file IDs sit between /d/ and /edit or in the ?id= param.
• Check mimeType before reading: Google-native files (application/vnd.google-apps.*) must be exported, not downloaded. Uploaded files (CSV, XLSX, PDF) use download.
• Spreadsheets: for a Google Sheet you need to read cell-by-cell or write back to, use a dedicated Google Sheets skill with the same file ID — export --mime-type text/csv only gives you a flat snapshot of the first tab.
• Bound your listings: pass --page-size on big folders and page through with --page-token instead of pulling everything at once.
• Shared drives: if a file the user can see is missing from results, retry with --all-drives.

Cross-Integration Workflows

• Drive → analysis: download --save-to a CSV/XLSX, then parse it with pandas and load it wherever it belongs.
• Any integration → Drive: pull data with the source CLI, write a CSV locally, then create-file --from-file into a Drive folder to share.
• Drive → Sheets: info a Google Sheet found in Drive, then hand its ID to a Google Sheets skill for structured reads and writes.

Important Notes

• Composio-powered: This integration uses Composio for OAuth and API execution. Tokens are managed automatically — you never see a Google refresh token in your env.
• Output: JSON to stdout. Commands return {"success": true, "data": {...}} (plus saved_to when --save-to is used) or {"success": false, "error": "...", "slug": "..."} with a non-zero exit code.
• Access scope: the agent only sees what the connected Google account can see. "Not found" usually means the file was never shared with that account, not that it doesn't exist.
• Large files: downloads land in the workspace filesystem — prefer exporting a CSV slice over pulling multi-hundred-MB originals.
• Safety: always confirm with the user before trashing, moving, overwriting, or sharing files — sharing changes who outside the org can read the data.