Skip to content

CLAUDE.md

Latest commit

 

History

History
114 lines (84 loc) · 2.95 KB

CLAUDE.md

File metadata and controls

114 lines (84 loc) · 2.95 KB
Raw

shieldcortex-sdk

TypeScript SDK for the ShieldCortex API — AI memory security scanning.

What This Package Does

This is a lightweight HTTP client that wraps the ShieldCortex SaaS API. It lets developers scan AI agent inputs/memory for prompt injection, credential leaks, encoding attacks, and other threats.

Installation

npm install shieldcortex-sdk

Quick Integration

import { ShieldCortex } from 'shieldcortex-sdk';

const sc = new ShieldCortex({ apiKey: process.env.SHIELDCORTEX_API_KEY! });

// Scan user input before processing
const result = await sc.scan({ content: userInput });
if (!result.allowed) {
  throw new Error(`Input blocked: ${result.firewall.reason}`);
}

Available Methods

Method Purpose
scan(input) Scan single content through 6-layer defence pipeline
scanBatch(items, opts?) Scan up to 100 items in one request
getAuditLogs(query?) Query audit trail with filters and pagination
getAuditStats(timeRange?) Get summary stats ('24h', '7d', '30d')
getQuarantine(query?) List quarantined items
reviewQuarantine(id, action) Approve or reject quarantined item

Key Types

// Constructor
new ShieldCortex({ apiKey: string, baseUrl?: string })

// Scan input
{ content: string, title?: string, source?: { type, identifier }, config?: { mode } }

// Scan result
{ allowed: boolean, firewall: { result, reason, anomalyScore }, trust: { score }, auditId: number }

// Defence modes: 'strict' | 'balanced' | 'permissive'
// Firewall results: 'ALLOW' | 'BLOCK' | 'QUARANTINE'

Error Classes

  • AuthError (401) — Invalid API key
  • RateLimitError (429) — Has .retryAfter property (seconds)
  • ValidationError (400) — Bad request body
  • ShieldCortexError — Base class, has .status and .body

Common Patterns

Middleware guard (Express/Hono)

async function shieldGuard(req, res, next) {
  const result = await sc.scan({ content: req.body.message });
  if (!result.allowed) {
    return res.status(403).json({ error: result.firewall.reason });
  }
  next();
}

LangChain tool wrapper

async function safeTool(input: string) {
  const check = await sc.scan({ content: input, source: { type: 'agent' } });
  if (!check.allowed) throw new Error('Blocked by ShieldCortex');
  return actualTool(input);
}

Batch processing

const results = await sc.scanBatch(
  messages.map(m => ({ content: m.text })),
  { config: { mode: 'strict' } }
);
const threats = results.results.filter(r => !r.allowed);

Project Structure

sdk/
  src/
    index.ts    — ShieldCortex client class
    types.ts    — All TypeScript interfaces
    errors.ts   — Error classes (AuthError, RateLimitError, etc.)
  dist/         — Compiled output

Development

npm run build    # Compile TypeScript

Zero runtime dependencies. Uses native fetch (Node 18+). ESM-only.