google-workspace（Google Workspace 办公协作）

Google Workspace CLI (gws)

One CLI for all of Google Workspace. Reads Google's Discovery Service at runtime, so new API endpoints are available automatically.

Prerequisites

• gws binary on $PATH (install: pnpm add -g @googleworkspace/cli)
• Authenticated: gws auth login (OAuth via browser)
• Auth status: gws auth status

Reference

All gws skill files are symlinked from the upstream Gemini extension at: ./gws-skills/

Key references:

• ./gws-skills/gws-shared/SKILL.md - Auth, global flags, security rules
• ./gws-skills/gws-drive/SKILL.md - Drive file/folder operations
• ./gws-skills/gws-docs/SKILL.md - Google Docs read/write
• ./gws-skills/gws-docs-write/SKILL.md - Append text to docs
• ./gws-skills/gws-sheets/SKILL.md - Sheets operations
• ./gws-skills/gws-gmail/SKILL.md - Gmail operations
• ./gws-skills/gws-calendar/SKILL.md - Calendar operations
• ./CONTEXT.md - Core syntax and usage patterns

Read the relevant skill file before executing commands you haven't used before.

Core Syntax

gws <service> <resource> [sub-resource] <method> [flags]

Common Flags

Schema Discovery

If you don't know the payload structure, check the schema first:

gws schema <service>.<resource>.<method>
# Example: gws schema drive.files.list
# Example: gws schema docs.documents.batchUpdate

Common Patterns

Drive - List/Search Files

gws drive files list --params '{"q": "name contains \"Report\"", "pageSize": 10}' --fields "files(id,name,mimeType)"

Drive - Create Folder

gws drive files create --json '{"name": "My Folder", "mimeType": "application/vnd.google-apps.folder"}'

Drive - Move File to Folder

gws drive files update --params '{"fileId": "FILE_ID", "addParents": "FOLDER_ID"}'

Drive - Export Google Doc as Plain Text

gws drive files export --params '{"fileId": "DOC_ID", "mimeType": "text/plain"}' --output /tmp/doc.txt

Docs - Read Document

gws docs documents get --params '{"documentId": "DOC_ID", "fields": "title,body"}'

Docs - Write to Document (batchUpdate)

For inserting/replacing text, build a JSON payload file to avoid shell escaping:

import json
payload = {
    "requests": [
        {"insertText": {"location": {"index": 1}, "text": "Hello world"}}
    ]
}
with open('/tmp/payload.json', 'w') as f:
    json.dump(payload, f)

gws docs documents batchUpdate --params '{"documentId": "DOC_ID"}' --json "$(cat /tmp/payload.json)"

Docs - Append Text (Helper)

gws docs +write --document DOC_ID --text 'Text to append'

Sheets - Read Cells

gws sheets spreadsheets values get --params '{"spreadsheetId": "ID", "range": "Sheet1!A1:C10"}'

Gmail - Search Messages

gws gmail users messages list --params '{"userId": "me", "q": "from:someone@example.com"}'

Agentic Workflow & Vibe Coding

• Iterative API Execution: Do not expect complex Workspace API payloads to succeed on the first try. Draft the payload, run it with the --dry-run flag, review the expected changes, isolate any schema errors, fix ONE field at a time, and re-test until the payload is valid before executing the actual mutation.
• Vibe Coding: Save your complex JSON payloads or script sequences to local files and commit them before running irreversible batch operations across Drive or Docs.

Rules

• Context window protection: ALWAYS use --fields on list/get to avoid massive JSON responses
• Dry-run first: Use --dry-run for destructive operations before executing
• Confirm writes: Ask the user before executing write/delete commands
• Shell escaping: For Sheets ranges with !, use single quotes to prevent bash history expansion
• Large payloads: Write JSON to a temp file and use $(cat /tmp/file.json) instead of inline