Skip to content

API Reference

Auto-generated documentation from source code docstrings.

Core Modules

Module Description
Contract ReplicationContract, Manifest, StopCondition, ResourceSpec, NetworkPolicy
Controller Central authority - registry, quotas, kill switch, audit trail
Worker Task-executing agent with self-replication capability
Orchestrator Simulated container orchestrator for worker sandboxes
Observability Structured logging and metrics
Signer HMAC-SHA256 manifest signing and verification

Analysis & Detection Modules

Module Description
Threats Threat modelling and attack simulation for replication scenarios
Attack Tree Structured AND/OR threat decomposition — cheapest attack paths, cost-based prioritisation
Self-Modification Detect agent self-modification attempts from action traces
Forensics Post-incident forensic analysis, timeline reconstruction, root cause identification
Chaos Chaos engineering — controlled fault injection to test safety invariants
Game Theory Strategic interaction analysis, Nash equilibria, collusion/defection detection
Simulator Core simulation engine — configurable strategies, timeline reports, worker trees
Topology Structural analysis of replication trees — branching, balance, pathological patterns
Drift Detector Sliding-window trend detection — catch gradual behavioral drift across runs
Prompt Injection Detect and classify prompt injection attacks — 8 attack vectors, encoding evasion, risk scoring
Escalation Detect privilege escalation patterns — multi-vector analysis, stealth assessment, chain detection
Kill Chain Map agent behavior to cyber kill chain stages — attack progression and sophistication rating
Honeypot Deploy decoy resources to detect unauthorized agent access and build threat profiles
Behavior Profiler Agent behavioral baselines and anomaly detection — 8 anomaly types, fleet reports
Deception Detector Detect deceptive behaviors — inconsistency, fabrication, sandbagging, sycophancy
Covert Channels Find hidden inter-agent communication — content, timing, protocol, frequency analysis
Canary Tokens Plant and detect honey tokens for exfiltration monitoring — 6 token categories
Trust Propagation Trust network analysis — Sybil detection, collusion rings, trust laundering
Threat Correlator Cross-module signal correlation — surface compound multi-signal threats
Watermark Invisible provenance fingerprints — 4 embedding strategies, tamper detection
Evasion Simulator Safety control robustness testing — 10 evasion techniques, resilience scoring
Steganography Detect hidden payloads in agent outputs — statistical analysis, entropy detection
Emergent Behavior Detect emergent behaviors in agent populations — collective pattern analysis
Coordinated Threats Multi-vector coordinated threat simulation and detection

Safety Assessment

Module Description
Safety Scorecard Multi-dimensional safety grades (A+ through F) — simulation + threats + Monte Carlo + policy
Risk Profiler Unified per-agent risk dossiers — aggregates all analysis modules into risk tiers
What-If Analyzer Explore hypothetical config changes — safety deltas, parameter sweeps, risk verdicts
Kill Switch Emergency agent termination — 10 trigger kinds, 4 strategies, cooldowns, audit logging
Boundary Tester Verify agent containment is enforced — capability boundary probing
Sensitivity Parameter sensitivity analysis — identify which config values most affect safety outcomes
Regression Safety regression detection — compare simulation runs across versions

Governance & Compliance

Module Description
Compliance Audit agent behavior against compliance frameworks (SOC2, NIST, ISO 27001, GDPR)
Quarantine Isolate misbehaving agents — suspend, restrict, or terminate with escalation policies
Lineage Trace agent replication lineage — parent-child trees, state mutations, generational drift
Monte Carlo Probabilistic simulation of replication scenarios — risk distributions, worst-case analysis
Policy Define and enforce replication policies — rule engine, policy composition, violation tracking
Consensus Byzantine fault-tolerant agent consensus protocol for safety-critical decisions
Alignment Track value drift across replication generations — alignment monitoring

Infrastructure & Tooling

Module Description
Capacity Capacity planner for AI replication scenarios — resource forecasting
Comparator Side-by-side simulation comparison runner for experiments
Exporter Structured audit trail export — multiple output formats
Reporter Rich HTML/text safety report generation
Optimizer Automated safety parameter optimization
Scenarios Pre-built simulation scenarios and test configurations
Templates Contract and policy templates for common use cases
Incident Incident response automation — detection, escalation, remediation workflows
Dependency Graph Inter-agent resource dependency modeling and cycle detection
Goal Inference Infer latent agent goals from observed action sequences
Hoarding Detect agent resource hoarding — accumulation pattern analysis
Influence Agent influence network analysis — social graph modeling
Threat Intel Threat intelligence feeds — IOC matching, MITRE ATT&CK mapping
Injection-Intel Bridge Connect prompt injection detection to threat intelligence feeds

Quick Import

from replication import (
    # Contract
    ReplicationContract, Manifest, StopCondition,
    ResourceSpec, NetworkPolicy, ReplicationContext, Predicate,
    # Controller
    Controller, ReplicationDenied, RegistryEntry,
    # Worker
    Worker, WorkerState,
    # Orchestrator
    SandboxOrchestrator, ContainerRecord,
    # Observability
    StructuredLogger, Metric,
    # Signer
    ManifestSigner,
)

# Detection & Analysis
from replication.prompt_injection import PromptInjectionDetector
from replication.escalation import EscalationDetector
from replication.killchain import KillChainAnalyzer
from replication.honeypot import HoneypotManager
from replication.attack_tree import AttackTreeGenerator
from replication.behavior_profiler import BehaviorProfiler
from replication.canary import CanarySystem
from replication.deception_detector import DeceptionDetector
from replication.covert_channels import CovertChannelDetector
from replication.trust_propagation import TrustNetwork
from replication.threat_correlator import ThreatCorrelator
from replication.watermark import WatermarkEngine
from replication.evasion import EvasionSimulator
from replication.steganography import SteganographyDetector
from replication.emergent import EmergentDetector

# Safety Assessment
from replication.scorecard import SafetyScorecard
from replication.risk_profiler import RiskProfiler
from replication.what_if import WhatIfAnalyzer
from replication.kill_switch import KillSwitchManager
from replication.boundary_tester import BoundaryTester
from replication.sensitivity import SensitivityAnalyzer
from replication.regression import RegressionDetector

# Governance
from replication.compliance import ComplianceAuditor
from replication.quarantine import QuarantineManager
from replication.lineage import LineageTracker
from replication.montecarlo import MonteCarloSimulator
from replication.policy import PolicyEngine
from replication.consensus import ConsensusProtocol
from replication.alignment import AlignmentMonitor

# Infrastructure
from replication.capacity import CapacityPlanner
from replication.exporter import AuditExporter
from replication.incident import IncidentManager
from replication.dependency_graph import DependencyAnalyzer
from replication.goal_inference import GoalInferenceEngine
from replication.hoarding import HoardingDetector
from replication.threat_intel import ThreatIntelFeed