T81 Policy Checker

T81 Policy Checker Skill

This skill evaluates Axion policies for T81 operations, providing deterministic policy validation and compliance checking.

Usage

When the user needs to validate operations against T81's Axion governance policies, use this skill to perform policy evaluation.

Commands

Check policy compliance:

check <operation> <resource> [--policy <policy_file>] [--context <context_file>]

Validate policy syntax:

validate-policy <policy_file> [--strict]

Test policy rules:

test-rule <policy_file> <rule_name> [--test-data <data_file>]

List available policies:

list-policies [--canonfs-root <path>]

Simulate policy decision:

simulate <operation> <resource> [--policy <policy_file>] [--dry-run]

Examples

# Check if import operation is allowed
check import model.t81w --policy security.apl

# Validate policy file syntax
validate-policy ./policies/secure_model.apl --strict

# Test specific rule with test data
test-rule security.apl allow_model_import --test-data test_cases.json

# List all policies in CanonFS
list-policies --canonfs-root ~/.t81_canonfs

# Simulate policy decision without execution
simulate inference neural_net.t81w --policy inference.apl --dry-run

Policy Operations

Supported Operations:

• import - Import files into CanonFS
• export - Export files from CanonFS
• execute - Run T81VM operations
• inference - Perform AI inference
• bundle_create - Create decision bundles
• bundle_consume - Consume decision bundles

Policy Context: Policy evaluation can include context from:

• User identity and permissions
• Resource metadata and hashes
• Temporal constraints and conditions
• Environment variables and system state

Output Format

Policy Allow:

{
  "decision": "allow",
  "operation": "import",
  "resource": "model.t81w",
  "policy": "security.apl",
  "reason": "approved_model_hash",
  "timestamp": "2026-04-04T15:30:00Z",
  "policy_hash": "CanonHash81..."
}

Policy Deny:

{
  "decision": "deny",
  "operation": "execute", 
  "resource": "untrusted_code.tisc",
  "policy": "security.apl",
  "reason": "unapproved_code_hash",
  "timestamp": "2026-04-04T15:30:00Z",
  "policy_hash": "CanonHash81...",
  "suggestions": [
    "Get code approved through security review",
    "Use approved code from trusted repository"
  ]
}

Policy Language Features

Axion Policy Syntax:

# Allow model inference for approved hashes
allow inference if model.hash in approved_models;

# Deny execution during maintenance windows
deny execute if time in maintenance_hours;

# Conditional resource access
allow export if user.role in ["admin", "auditor"] 
          and resource.classification <= "secret";

Policy Evaluation:

• Deterministic rule evaluation
• No runtime ambiguity or heuristics
• Complete audit trail for decisions
• Cryptographic policy verification

Error Handling

• Clear error messages for policy syntax issues
• Detailed explanations for policy denials
• Suggestions for policy remediation
• Validation of policy file integrity

Security Notes

• Policy files are stored immutably in CanonFS
• Policy evaluation is deterministic and reproducible
• All policy decisions are logged for audit
• Policy changes require proper governance

Integration

This skill wraps T81's Axion policy engine and provides standardized access to policy validation APIs.