@sentinel-atl/sdk

Developer-friendly SDK for the Agent Trust Layer — add trust to your agent in ~5 lines.

Features

• Single entry point — createTrustedAgent() wires up identity, audit, reputation, revocation, offline, and safety
• Offline mode — goOffline() / goOnline() with cached trust decisions and CRDT merge
• Content safety — checkSafety() with pluggable classifiers (regex, keyword, custom)
• Re-exports all core types for convenience

Install

npm install @sentinel-atl/sdk

Quick Start

import { createTrustedAgent } from '@sentinel-atl/sdk';

const agent = await createTrustedAgent({
  label: 'my-agent',
  enableSafety: true,
});

// Issue a credential
const vc = await agent.issueVC({
  subjectDid: 'did:key:z6Mk...',
  scope: ['files:read'],
  maxDelegationDepth: 0,
});

// Verify trust
const result = await agent.verifyVC(vc);

// Check content safety
const safety = await agent.checkSafety('user input text');

// Go offline with cached decisions
agent.goOffline();
const decision = agent.evaluateTrust(vc);
agent.goOnline();

API

| Method | Description | |---|---| | createTrustedAgent(config) | Factory — returns a TrustedAgent with all subsystems | | agent.issueVC(options) | Issue a Verifiable Credential | | agent.verifyVC(vc) | Verify a VC | | agent.vouch(targetDid, score, evidence) | Submit a reputation vouch | | agent.getReputation(did) | Get reputation score | | agent.checkSafety(content) | Run content safety pipeline | | agent.goOffline() / goOnline() | Toggle offline mode | | agent.evaluateTrust(vc) | Evaluate trust using cached data | | agent.mergeReputationState(remote) | CRDT merge remote reputation | | agent.revokeVC(vcId, reason) | Revoke a credential | | agent.backup() / recover(shares) | Shamir key backup/recovery |

License

MIT