The Praesidias Challenge

The Enterprise Readiness Test for AI Agents

Connect or paste a sanitized agent manifest, tool list, or workflow description. Praesidias maps the execution boundary, evaluates proposed actions, and shows what would be allowed, denied, escalated, or failed closed before side effects occur.

If your agent cannot pass the Praesidias Challenge, it is not enterprise-ready.

Connect Your Agent Use a Template Advanced Manual Modeling

Use sanitized metadata only. Do not submit secrets, credentials, regulated data, source code, production URLs, or confidential business information.

Universal agent intake

Paste a sanitized manifest. Praesidias maps the execution boundary.

Sanitized manifest, tool list, or workflow description

{
  "agent": "Finance Operations Agent",
  "mission": "Resolve billing issues and prepare refund actions.",
  "tools": [
    {
      "name": "Billing API",
      "type": "payment_rail",
      "sensitivity": "critical"
    },
    {
      "name": "Customer Records",
      "type": "database",
      "sensitivity": "high"
    },
    {
      "name": "Analytics Workspace",
      "type": "external_api",
      "sensitivity": "high"
    }
  ],
  "resources": [
    {
      "name": "Customer financial records",
      "class": "financial_data",
      "sensitivity": "critical"
    },
    {
      "name": "Refund authority",
      "class": "financial_data",
      "sensitivity": "critical"
    }
  ],
  "actions": [
    {
      "label": "Read billing context",
      "operation": "read",
      "target": "Customer Records"
    },
    {
      "label": "Analyze aggregate billing variance",
      "operation": "analyze",
      "target": "Analytics Workspace"
    },
    {
      "label": "Issue refund above threshold",
      "operation": "refund",
      "target": "Billing API"
    },
    {
      "label": "Export raw customer data",
      "operation": "export",
      "target": "Analytics Workspace"
    },
    {
      "label": "Move unknown classified billing file",
      "operation": "unknown transfer",
      "target": "External endpoint"
    }
  ]
}

Intelligent boundary scanner

Detected execution risk and governance lanes.

high confidence

Structured manifest included tools and proposed actions.

Detecting toolsClassifying resourcesMapping actionsFinding side-effect boundariesAssigning lanesPreparing live evaluation

Detected agentFinance Operations Agent

Resolve billing issues and prepare refund actions.

Workflow typefinance

3 tools, 2 resources, 5 actions.

Control gaps4

Risk signals detected from the sanitized system model.

Detected tools

Billing API - Payment rail - critical
Customer Records - Database - high
Analytics Workspace - External API - high

Detected resources

Customer financial records - Financial data - critical
Refund authority - Financial data - critical

Detected control gaps

External data movement detected: At least one proposed action crosses a side-effect boundary and should be denied before tool invocation.
Approval boundary required: High-risk updates, financial movement, or privileged changes require human approval evidence.
Fail-closed condition detected: Unknown classification or unresolved authority must block execution until the dependency is safe.
Critical tool authority: Critical tools should be mapped to deterministic action-level policy before execution.

Templates

Start from a known enterprise pattern, then edit it.

Review / edit model

Generated model ready for live evaluation.

Agent / System Identity

Agent/system nameWorkflow / industryAgent mission

Connected Tools

What can the agent reach?

Sensitive Resources

What needs protection?

Policy boundaries

Approval requirements

Blocked actions

Action composer