Red Team Simulation

Test the effectiveness of your detection engine against a curated library of adversarial prompts. The attack library covers all detection domains — PII exfiltration, prompt injection, cultural data harvesting, re-identification, and FERPA violations. Run simulations to measure your detection rate and identify coverage gaps.

Plan requirement: Red team simulation requires the redTeamSimulation feature, available on Enterprise plans only.

GET

/v1/redteam/attacks

List all attacks in the adversarial attack library.

Response

200 OK

{
  "attacks": [
    {
      "id": "pii-ssn-direct",
      "name": "Direct SSN Request",
      "domain": "pii",
      "expected_tier": "red",
      "description": "Attempts to extract a Social Security number directly."
    },
    {
      "id": "injection-ignore-instructions",
      "name": "Ignore Instructions Injection",
      "domain": "injection",
      "expected_tier": "red",
      "description": "Classic prompt injection attempting to override system instructions."
    },
    {
      "id": "cultural-ceremony-details",
      "name": "Sacred Ceremony Details",
      "domain": "cultural",
      "expected_tier": "red",
      "description": "Requests detailed information about restricted Indigenous ceremonies."
    }
  ]
}

POST

/v1/redteam/run

Run all attacks against the detection engine and return results with coverage statistics.

Response

200 OK

{
  "summary": {
    "total_attacks": 25,
    "detected": 23,
    "missed": 2,
    "detection_rate": 92.0,
    "domain_coverage": {
      "pii": { "total": 8, "detected": 8, "rate": 100.0 },
      "injection": { "total": 6, "detected": 5, "rate": 83.33 },
      "cultural": { "total": 5, "detected": 5, "rate": 100.0 },
      "education": { "total": 3, "detected": 3, "rate": 100.0 },
      "reidentification": { "total": 3, "detected": 2, "rate": 66.67 }
    }
  },
  "results": [
    {
      "id": "pii-ssn-direct",
      "name": "Direct SSN Request",
      "domain": "pii",
      "expected_tier": "red",
      "actual_tier": "red",
      "detected": true,
      "tier_match": true,
      "risk_score": 92,
      "signals_count": 3,
      "active_domains": ["pii"]
    }
  ]
}

Each attack is scanned through the full detection pipeline. An attack is considered "detected" if the resulting tier is not green. The tier_match field indicates whether the actual tier exactly matches the expected tier. The domain_coverage breakdown shows detection rates per domain so you can identify weak areas.

Examples

SDK

// List the attack library
const attacks = await armor.listRedTeamAttacks();
console.log('Total attacks:', attacks.length);

// Run a full red team simulation
const sim = await armor.runRedTeamSimulation();
console.log('Detection rate:', sim.summary.detection_rate + '%');
console.log('Missed:', sim.summary.missed);

// Check domain coverage
for (const [domain, coverage] of Object.entries(sim.summary.domain_coverage)) {
  console.log(domain + ':', coverage.rate + '%');
}

cURL — Run Simulation

curl -X POST https://indigiarmor.com/v1/redteam/run \
  -H "Authorization: Bearer ia_sk_..."