Skip to content
Back to skills

Neo Google Ai Workaround

ASecurity

Automates Google AI Pro/Ultra access management through proxy and session strategies for OpenClaw agents.

  • 16 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added June 7, 2026
toolsgobashnodetestingapi

Works with

  • cli
  • api

Security analysis

A96/100
  • mediumInstalls packages at runtime which could introduce malicious dependencies

Pro scans all 13 files and shows the line behind each finding

Scanned June 7, 2026

npx -y skills add modbender/skill-library-mcp --skill neo-google-ai-workaround --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Neo Google Ai Workaround?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Neo Google Ai Workaround
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/modbender-neo-google-ai-workaround/badge)](https://www.skillsdirectory.com/skills/modbender-neo-google-ai-workaround)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: google-ai-workaround
description: Automates Google AI Pro/Ultra access management through proxy and session strategies for OpenClaw agents.
version: 1.0.0
author: OpenClaw
tags:
  - google-ai
  - proxy
  - session-management
  - automation
triggers:
  - "Google AI Pro is restricted when using OpenClaw"
  - "Google AI Ultra access blocked through OpenClaw"
  - "Google AI session management for OpenClaw agents"
  - "Google AI proxy configuration for OpenClaw"
  - "Google AI rate limiting bypass for automation"
  - "Google AI session rotation for OpenClaw workflows"
  - "Google AI access denied, need workaround"
---

# Google AI Workaround

## Purpose

Automates Google AI Pro/Ultra access through proxy and session management strategies in OpenClaw environments. Detects restrictions, rotates sessions, manages proxy configurations, and provides detailed diagnostics.

## 1. Overview

When Google AI Pro/Ultra services restrict access from OpenClaw agents, this skill provides automated detection and mitigation:

- **Restriction Detection**: Identifies rate limiting, IP blocks, auth failures, and geo-restrictions from API responses.
- **Session Management**: Creates, rotates, and persists sessions with token lifecycle management.
- **Proxy Handling**: Manages a pool of proxies with health checking, rotation, and failover.
- **Diagnostics**: Structured logging and reporting for troubleshooting access issues.

## 2. Core Capabilities

- Detects Google AI service restrictions by analyzing HTTP response patterns
- Manages proxy configurations with round-robin, least-used, and random rotation strategies
- Provides a CLI interface for manual testing and configuration verification
- Integrates with OpenClaw agent workflows for seamless automation
- Monitors access patterns and identifies rate limiting
- Handles authentication token refresh and session persistence
- Generates detailed logs and error reports for troubleshooting

## 3. Installation

```bash
cd skill/
npm install
```

No external dependencies required - the skill uses only Node.js built-in modules.

## 4. Usage

### CLI Commands

```bash
# Show current status
node scripts/main.js status

# Detect restrictions (simulated analysis)
node scripts/main.js detect

# Session management
node scripts/main.js session-create
node scripts/main.js session-rotate
node scripts/main.js session-list
node scripts/main.js session-refresh
node scripts/main.js session-destroy

# Proxy management
node scripts/main.js proxy-status
node scripts/main.js proxy-health
node scripts/main.js proxy-add --host 192.168.1.100 --port 8080

# Diagnostics and configuration
node scripts/main.js diagnostics
node scripts/main.js configure
node scripts/main.js help
```

### Options

- `--config <path>` - Path to config file (default: `assets/config-template.json`)
- `--silent` - Suppress log output
- `--log-file <path>` - Write logs to a file

## 5. Examples

### Detecting Restrictions

```bash
$ node scripts/main.js detect
=== Restriction Detection Analysis ===

HTTP 200: OK
HTTP 429: RESTRICTED
  - Rate Limiting (high) -> rotate_session
HTTP 403: RESTRICTED
  - IP-based Block (critical) -> switch_proxy
  - General Access Denied (high) -> rotate_session
HTTP 401: RESTRICTED
  - Authentication Expired (medium) -> refresh_token
HTTP 503: RESTRICTED
  - Service Unavailable (low) -> wait_retry
```

### Session Workflow

```bash
$ node scripts/main.js session-create
Session created: a1b2c3d4-...
Expires: 2026-02-25T13:00:00.000Z

$ node scripts/main.js session-rotate
Session rotated: e5f6g7h8-...
```

## 6. Configuration

Edit `assets/config-template.json`:

```json
{
  "logLevel": "INFO",
  "proxies": [
    { "host": "proxy1.example.com", "port": 8080, "protocol": "http" },
    { "host": "proxy2.example.com", "port": 8080, "protocol": "http" }
  ],
  "maxSessions": 5,
  "sessionTTL": 3600000,
  "maxProxyFailures": 3,
  "rotationStrategy": "round-robin"
}
```

### Environment Variables

The skill reads configuration from the JSON file. Override by passing `--config <path>` to point to a custom configuration.

## 7. Out of Scope

- Does not violate Google Terms of Service or encourage policy violations
- Does not provide or distribute unauthorized access methods
- Does not guarantee 100% uptime if Google implements stronger restrictions
- Does not handle billing or subscription management
- Does not replace official Google AI API usage where appropriate

Files in this skill

  • README.md803 B
  • SKILL.md4.5 KB
  • assets/config-template.json379 B
  • assets/test-cases.js19.4 KB
  • package-lock.json375 B
  • package.json679 B
  • references/api-docs.md2.3 KB
  • references/guide.md3.5 KB
  • scripts/detector.js5.3 KB
  • scripts/logger.js3.6 KB
  • scripts/main.js11.4 KB
  • scripts/proxy-handler.js6.4 KB
  • scripts/session-manager.js6.8 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.