← All learning paths

Learning path · 5 steps

Validate agent trust boundaries

Review tool authority, gateway enforcement, adversarial experiments and detection evidence together.

Agent builders & security evaluators

  1. 01

    Advanced guide

    Engineer MCP tool trust boundaries beyond the prompt

    Bind tool calls and approvals to the intended identity and action.

    Open guide →
  2. 02

    Advanced guide

    Design AI gateway policies that preserve security boundaries

    Test the controls enforced at each request boundary.

    Open guide →
  3. 03

    Advanced guide

    Build adversarial AI evaluations with defensible evidence

    Measure observed outcomes and separate grader judgments from evidence.

    Open guide →
  4. 04

    Advanced guide

    Engineer Sentinel correlations that survive real telemetry

    Document identity joins, event-time windows and rule limitations.

    Open guide →
  5. 05

    Advanced Sentinel hunt

    Compare identity and Azure control-plane ingestion health

    Inspect source-specific delay and query coverage before alerting.

    Open query →