ASI-02: Insufficient Authorization

How MandateZ mitigates OWASP Agentic Security ASI-02 — Insufficient Authorization

How MandateZ Fixes ASI-02: Insufficient Authorization

What Is the Risk?

Insufficient Authorization happens when an AI agent can perform actions without proper permission checks. The agent may inherit the user’s full session token, bypass role-based access controls, or escalate privileges by chaining tool calls — all without any authorization layer verifying whether that specific agent should be allowed to act.

Most agent frameworks today have zero authorization enforcement at the agent level.

How MandateZ Mitigates It

MandateZ adds a dedicated authorization layer between the agent and every resource it touches.

Policy-Based Access Control

Every action an agent attempts is evaluated against its assigned policy before execution. Policies are explicit rules — not prompt instructions an LLM can ignore.

import { MandateZClient, generateAgentIdentity } from '@mandatez/sdk';

const identity = await generateAgentIdentity();

const client = new MandateZClient({
  agentId: identity.agent_id,
  ownerId: 'your_org_id',
  privateKey: identity.private_key,
  supabaseUrl: process.env.SUPABASE_URL!,
  supabaseAnonKey: process.env.SUPABASE_ANON_KEY!,
  policies: [{
    id: 'pol_auth',
    owner_id: 'your_org_id',
    name: 'Finance Bot Authorization',
    rules: [
      // Can read invoices
      { id: 'r1', action_types: ['read'], resource_pattern: 'invoices/*', effect: 'allow' },
      // Can call the billing API
      { id: 'r2', action_types: ['call'], resource_pattern: 'api/billing', effect: 'allow' },
      // Cannot touch anything else
      { id: 'r3', action_types: ['read', 'write', 'delete', 'export', 'call', 'payment'], resource_pattern: '*', effect: 'block' },
    ],
  }],
});

// Attempt to read HR records — blocked by policy
const event = await client.track({
  action_type: 'read',
  resource: 'hr/employee_salaries',
});

console.log(event.outcome);   // 'blocked'
console.log(event.policy_id); // 'pol_auth'

Cryptographic Agent Identity

Each agent gets its own Ed25519 keypair. Actions are signed with the agent’s private key, making it impossible for one agent to impersonate another or act without a verifiable identity.

import { generateAgentIdentity, verifyEvent } from '@mandatez/sdk';

const identity = await generateAgentIdentity();
// Each agent has a unique keypair — no shared tokens
console.log(identity.agent_id);   // 'ag_V1StGXR8_Z5jdHi6B-myT'
console.log(identity.public_key); // Ed25519 public key

// Any event can be verified against the agent's public key
const valid = await verifyEvent(event);
console.log(valid); // true — proves this agent authored this action

Row-Level Security

On the data layer, Supabase RLS ensures each owner only sees their own agents and events. There is no API endpoint that returns cross-tenant data.

---

Get Started

Set up MandateZ in under 5 minutes and add real authorization to your agents.