@surfinguard/core-engine-wasm

WASM-powered heuristic scoring engine for the Surfinguard AI Security SDK. Provides 18 analyzers and 152 threat patterns compiled to WebAssembly for near-native performance.

Most users should install @surfinguard/sdk instead. This package is the low-level engine used internally by the SDK, CLI, and MCP server.

Install

npm install @surfinguard/core-engine-wasm

Usage

import { initialize, CoreEngine } from '@surfinguard/core-engine-wasm';

// Must initialize WASM before first use
await initialize();

const engine = new CoreEngine();
const result = engine.check({ type: 'url', value: 'https://example.com' });

console.log(result.level);  // 'SAFE' | 'CAUTION' | 'DANGER'
console.log(result.score);  // 0-10

Supported Action Types

url, command, text, file_read, file_write, api_call, query, code, message, transaction, auth, git, ui_action, infra, agent_comm, data_pipeline, document, iot

Risk Primitives

Each action is scored against 5 risk primitives: DESTRUCTION, EXFILTRATION, ESCALATION, PERSISTENCE, MANIPULATION. The composite score is max(primitive_scores).

| Level | Score | |-------|-------| | SAFE | 0–2 | | CAUTION | 3–6 | | DANGER | 7–10 |

License

MIT