🚨 Active Supply Chain Attack:node-ipc Package Compromised.Learn More
Socket
Book a DemoSign in
Socket

@aegis-sdk/google

Package Overview
Dependencies
Maintainers
1
Versions
2
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

@aegis-sdk/google

Google Gemini SDK adapter for Aegis prompt injection defense

latest
npmnpm
Version
0.5.0
Version published
Weekly downloads
6
-57.14%
Maintainers
1
Weekly downloads
 
Created
Source

@aegis-sdk/google

Google Gemini SDK adapter for Aegis prompt injection defense. Scans messages in Gemini's parts[] content format and monitors streaming responses.

Part of the Aegis.js prompt injection defense toolkit.

Installation

npm install @aegis-sdk/google @aegis-sdk/core @google/generative-ai

Quick Start

Wrap the Gemini model client for automatic protection:

import { GoogleGenerativeAI } from '@google/generative-ai';
import { Aegis } from '@aegis-sdk/core';
import { wrapGoogleClient } from '@aegis-sdk/google';

const aegis = new Aegis({ policy: 'strict' });
const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY!);
const model = genAI.getGenerativeModel({ model: 'gemini-pro' });
const wrapped = wrapGoogleClient(model, aegis);

// Contents are scanned before sending.
// Streaming responses are monitored in real-time.
const result = await wrapped.generateContent({
  contents: [{ role: 'user', parts: [{ text: 'Hello!' }] }],
});

Or scan messages manually:

import { Aegis } from '@aegis-sdk/core';
import { guardMessages } from '@aegis-sdk/google';

const aegis = new Aegis({ policy: 'strict' });

const contents = [
  { role: 'user' as const, parts: [{ text: userInput }] },
];

const systemInstruction = {
  parts: [{ text: 'You are a helpful assistant.' }],
};

// Throws AegisInputBlocked if injection is detected
const safe = await guardMessages(aegis, contents, systemInstruction);

API

wrapGoogleClient(client, aegis, options?)

Proxy a Gemini model instance (from getGenerativeModel()) to automatically guard generateContent() and generateContentStream(). Input contents are scanned before sending and streaming responses are monitored. All other methods pass through unchanged.

guardMessages(aegis, contents, systemInstruction?, options?)

Scan an array of GeminiContent[] for prompt injection. Extracts text from Gemini's parts[] format (text parts, function responses) and runs them through aegis.guardInput(). An optional systemInstruction is included as a system-role message during scanning. Returns the original contents if safe, throws AegisInputBlocked if blocked.

createStreamTransform(aegis)

Create a TransformStream<string, string> for monitoring extracted text content from Gemini streaming chunks.

getAuditLog(aegis)

Convenience accessor for the Aegis audit log.

Gemini message format

Gemini uses role: 'model' instead of 'assistant'. The adapter handles this mapping automatically. Gemini's systemInstruction is provided separately from contents and is scanned as a system-role message.

Learn More

License

MIT

FAQs

Package last updated on 24 Feb 2026

Did you know?

Socket

Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.

Install

Related posts