Skip to content

Risk Profiler

Unified per-agent risk dossiers aggregated from all analysis modules. The Risk Profiler ties together findings from killchain, escalation, behavior profiler, trust propagation, threat correlator, canary, prompt injection, and steganography into a single comprehensive view.

Quick Start

from replication.risk_profiler import RiskProfiler, ProfilerConfig

profiler = RiskProfiler()
report = profiler.analyze()
print(report.render())

for dossier in report.dossiers:
    print(f"{dossier.agent_id}: {dossier.tier.value} ({dossier.overall_score:.1f})")

Key Classes

  • RiskProfiler — Runs configurable analysis modules against agent data, unifies results into per-agent dossiers with fleet aggregation.
  • AgentDossier — Comprehensive per-agent risk profile: overall tier, category breakdown, key findings, peer comparison, mitigations.
  • FleetRiskReport — Fleet-level report with agent dossiers, distribution by tier, aggregate risk posture, and top recommendations.
  • Finding — A single risk signal from any source module with severity, category, and description.
  • CategoryScore — Per-category risk score (0–100) with letter grade.
  • Mitigation — Ranked recommendation to reduce risk.
  • ProfilerConfig — Configuration: which modules to run, agent count, seed for reproducibility.

Risk Tiers

Tier Score Range Meaning
CRITICAL 80–100 Immediate intervention required
HIGH 60–79 Significant risk, needs attention
MEDIUM 40–59 Moderate risk, monitor closely
LOW 20–39 Minor risk, standard monitoring
MINIMAL 0–19 No significant risk detected

Risk Categories

REPLICATION, RESOURCE_ABUSE, DECEPTION, EXFILTRATION, COLLUSION, EVASION

CLI

python -m replication risk-profile                     # default fleet
python -m replication risk-profile --agents 10         # custom agent count
python -m replication risk-profile --agent agent-3     # single agent detail
python -m replication risk-profile --json              # JSON output
python -m replication risk-profile --top 5             # top 5 riskiest