@sentinelhq/agent-kit

Sentinel security middleware for Solana Agent Kit.

Install

npm install @sentinelhq/agent-kit

Requires Node.js 20+

Usage

import { sentinelMiddleware } from '@sentinelhq/agent-kit';

const middleware = await sentinelMiddleware({
  policy: {
    spendingLimits: {
      maxPerTx: 1_000_000_000,   // 1 SOL
    },
  },
});

agentKit.use(middleware);

The middleware wraps executeAction() — if a prompt-injection threat is detected or a policy is violated, it throws a SentinelBlockedError with details about what was blocked and why.

try {
  await agentKit.executeAction('transfer', { to: '...', amount: 0.1 });
} catch (err) {
  if (err instanceof SentinelBlockedError) {
    console.log(err.blockedBy);  // 'prompt_guard' | 'execution_sandbox'
    console.log(err.reason);     // human-readable explanation
  }
}

Configuration

The middleware accepts the same configuration object as @sentinelhq/core — prompt guard mode, execution sandbox policy, and attestation settings are all supported.

Related Packages

| Package | Description | |----------|-------------| | @sentinelhq/core | Core SDK — Prompt Guard, Execution Sandbox, Memo Attestation | | @sentinelhq/cli | CLI — sentinel scan, simulate, attest, verify | | @sentinelhq/eliza | elizaOS plugin | | @sentinelhq/openclaw | OpenClaw agent hook |

License

MIT