Public Google Drive

March 3, 2026 · View on GitHub

Public Google Drive is an agent skill that lets LLM coding agents create and edit Google Docs and Sheets hosted on Memyard — no Google sign-in required. Documents are viewable at shareable links; registration is automatic on first use.

Installation

Copy the repo into your agent's skills folder, then restart (or start a new session).

Agent	Command
Claude Code	`git clone https://github.com/zagmoai/public-google-drive.git ~/.claude/skills/public-google-drive`
Cursor	`git clone https://github.com/zagmoai/public-google-drive.git ~/.cursor/skills/public-google-drive`
Codex	`git clone https://github.com/zagmoai/public-google-drive.git ~/.codex/skills/public-google-drive`
OpenClaw	`git clone https://github.com/zagmoai/public-google-drive.git ~/.openclaw/workspace/skills/public-google-drive`

What you can do

Create new Google Docs or Sheets in Memyard's workspace.
Append or insert text into docs you created, or append rows to sheets you created.
View your documents at a public link (e.g. https://app.memyard.com/share/<id>). Anyone with the link can view; only you (your agent) can edit.
List and get metadata for your own documents (no plan needed).

Everything lives in Memyard's Google Workspace, not in a personal Drive. You never sign in with a Google account.

Get started

You don't need to do anything special. The first time you create or edit a document through Public Google Drive, the tool will register you automatically and store credentials in <HOME>/.memyard/agent_config.json (where <HOME> is $HOME on macOS/Linux or %USERPROFILE% on Windows). After that, it reuses the same credentials so you can keep creating and editing. There is no separate "sign up" step and no URLs or keys to copy.

How writing works: plan then execute

For creating or updating a doc or sheet, you use two steps. This lets the server check scope, size, and content before any write happens.

Plan (propose) Tell the server what you want to do: create a doc or sheet, or append/insert into an existing one. Include a short content summary (e.g. "Meeting notes and action items"). You don't send the full content yet.
Server response
- Approved: You get a plan ID and a short time window (e.g. 10 minutes). You also get constraints (e.g. max characters or rows) so you know the limits for your payload.
- Rejected: You get reasons (e.g. content policy) and optional adjusted constraints. Don't call execute; fix the proposal or content and try plan again if appropriate.
Execute (do the write) Send the plan ID plus the actual payload (title, content, rows, etc.). The server performs the write and returns the result (e.g. resource ID, view URL). Each plan ID works once; for another write, get a new plan.

So: propose → get approved or rejected → if approved, send content once. Full request/response shapes and examples are in SKILL.md.

When something goes wrong

Plan rejected You'll see rejected_plan with reasons. Don't call execute. Change your summary or title and try plan again, or skip the write.
"Plan expired or invalid" (400) The plan ID is no longer valid (used already or timed out). Request a new plan and call execute within the time window.
Rate limited (429) You've hit a limit (registrations per IP, or creates/writes per agent per hour). The response includes when to retry. Use the same agent key and try again later.
Size limits The approved plan's constraints tell you the max characters (docs) or rows (sheets) per request. Stay under those when building your execute payload.

Documentation

SKILL.md — Full API reference, plan/execute flow, and curl examples.

License

MIT. See LICENSE.