Google Apps ScriptSAFE

Skills for Claude Code CLI such as full stack dev Cloudflare, React, Tailwind v4, and AI integrations.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: google-apps-script
description: "Build Google Apps Script automation for Sheets and Workspace. Custom menus, triggers (onEdit / time-driven / form submit), dialogs, sidebars, email batches, PDF export, external API. Use whenever the user wants to automate a Google Sheet, build a Sheets menu / sidebar / dialog, hit a Sheets row from email or a webhook, schedule a Sheets workflow, or asks 'how do I script this in Sheets'."
compatibility: claude-code-only
---

# Google Apps Script

Build automation scripts for Google Sheets and Workspace apps. Scripts run server-side on Google's infrastructure with a generous free tier.

## What You Produce

- Apps Script code pasted into Extensions > Apps Script
- Custom menus, dialogs, sidebars
- Automated triggers (on edit, time-driven, form submit)
- Email notifications, PDF exports, API integrations

## Workflow

### Step 1: Understand the Automation

Ask what the user wants automated. Common scenarios:
- Custom menu with actions (report generation, data processing)
- Auto-triggered behaviour (on edit, on form submit, scheduled)
- Sidebar app for data entry
- Email notifications from sheet data
- PDF export and distribution

### Step 2: Generate the Script

Follow the structure template below. Every script needs a header comment, configuration constants at top, and `onOpen()` for menu setup.

### Step 3: Provide Installation Instructions

All scripts install the same way:
1. Open the Google Sheet
2. **Extensions > Apps Script**
3. Delete any existing code in the editor
4. Paste the script
5. Click **Save**
6. Close the Apps Script tab
7. **Reload the spreadsheet** (onOpen runs on page load)

### Step 4: First-Time Authorisation

Each user gets a Google OAuth consent screen on first run. For unverified scripts (most internal scripts), users must click:

**Advanced > Go to [Project Name] (unsafe) > Allow**

This is a one-time step per user. Warn users about this in your output.

---

## Script Structure Template

Every script should follow this pattern:

```javascript
/**
 * [Project Name] - [Brief Description]
 *
 * [What it does, key features]
 *
 * INSTALL: Extensions > Apps Script > paste this > Save > Reload sheet
 */

// --- CONFIGURATION ---
const SOME_SETTING = 'value';

// --- MENU SETUP ---
function onOpen() {
  const ui = SpreadsheetApp.getUi();
  ui.createMenu('My Menu')
    .addItem('Do Something', 'myFunction')
    .addSeparator()
    .addSubMenu(ui.createMenu('More Options')
      .addItem('Option A', 'optionA'))
    .addToUi();
}

// --- FUNCTIONS ---
function myFunction() {
  // Implementation
}
```

---

## Critical Rules

### Public vs Private Functions

Functions ending with `_` (underscore) are **private** and CANNOT be called from client-side HTML via `google.script.run`. This is a silent failure -- the call simply doesn't work with no error.

```javascript
// WRONG - dialog can't call this, fails silently
function doWork_() { return 'done'; }

// RIGHT - dialog can call this
function doWork() { return 'done'; }
```

**Also applies to**: Menu item function references must be public function names as strings.

### Batch Operations (Critical for Performance)

Read/write data in bulk, never cell-by-cell. The difference is 70x.

```javascript
// SLOW (70 seconds on 100x100) - reads one cell at a time
for (let i = 1; i <= 100; i++) {
  const val = sheet.getRange(i, 1).getValue();
}

// FAST (1 second) - reads all at once
const allData = sheet.getRange(1, 1, 100, 1).getValues();
for (const row of allData) {
  const val = row[0];
}
```

Always use `getRange().getValues()` / `setValues()` for bulk reads/writes.

### V8 Runtime

V8 is the **only** runtime (Rhino was removed January 2026). Supports modern JavaScript: `const`, `let`, arrow functions, template literals, destructuring, classes, async/generators.

**NOT available** (use Apps Script alternatives):

| Missing API | Apps Script Alternative |
|-------------|------------------------|
| `setTimeout` / `setInterval` | `Utilities.sleep(ms)` (blocking) |
| `fetch` | `UrlFetchApp.fetch()` |
| `FormData` | Build payload manually |
| `URL` | String manipulation |
| `crypto` | `Utilities.computeDigest()` / `Utilities.getUuid()` |

### Flush Before Returning

Call `SpreadsheetApp.flush()` before returning from functions that modify the sheet, especially when called from HTML dialogs. Without it, changes may not be visible when the dialog shows "Done."

### Simple vs Installable Triggers

| Feature | Simple (`onEdit`) | Installable |
|---------|-------------------|-------------|
| Auth required | No | Yes |
| Send email | No | Yes |
| Access other files | No | Yes |
| URL fetch | No | Yes |
| Open dialogs | No | Yes |
| Runs as | Active user | Trigger creator |

Use simple triggers for lightweight reactions. Use installable triggers (via `ScriptApp.newTrigger()`) when you need email, external APIs, or cross-file access.

### Custom Spreadsheet Functions

Functions used as `=MY_FUNCTION()` in cells have strict limitations:

```javascript
/**
 * Calculates something custom.
 * @param {string} input The input value
 * @return {string} The result
 * @customfunction
 */
function MY_FUNCTION(input) {
  // Can use: basic JS, Utilities, CacheService
  // CANNOT use: MailApp, UrlFetchApp, SpreadsheetApp.getUi(), triggers
  return input.toUpperCase();
}
```

- Must include `@customfunction` JSDoc tag
- 30-second execution limit (vs 6 minutes for regular functions)
- Cannot access services requiring authorisation

---

## Quotas and Limits

| Resource | Free Account | Google Workspace |
|----------|-------------|-----------------|
| Script runtime | 6 min / execution | 6 min / execution |
| Time-driven trigger runtime | 30 min | 30 min |
| Triggers total daily runtime | 90 min | 6 hours |
| Triggers total | 20 per user per script | 20 per user per script |
| Email recipients/day | 100 | 1,500 |
| URL Fetch calls/day | 20,000 | 100,000 |
| Properties storage | 500 KB | 500 KB |
| Custom function 

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (1)

// On edit with full permissions (can send email, fetch URLs)

Gates applied: no_behavioural_pass.