Technical Reference

Deep dives on controls, infrastructure, operations, and metrics. Grouped by purpose so you can find the page you need without scanning the whole list.

Judge Internals

Building, selecting, validating, and operating the judge model.

Document	Description
Model-as-Judge Implementation	Detailed judge implementation, including prompt structure and scoring.
Judge Model Selection	Selection principles: family diversity, cost, latency, safety posture.
Judge Precedents	Building a precedent library so judge decisions stay consistent.
Distilling the Judge into an SLM	Moving from a large-model async judge to an inline sidecar SLM.
Output Evaluator	Output-side evaluation patterns that complement judge rulings.

Detection and SOC

Integrating AI runtime telemetry into security operations.

Document	Description
SOC Integration	SOC architecture, alert taxonomy, triage procedures.
SOC Content Pack	SIEM detection rules, correlation searches, dashboards.
Anomaly Detection Ops	Behavioural anomaly detection, operated as a programme.
Behavioral Anomaly Detection	Detection techniques and baselines for agent behaviour.
Graph-Based Agent Monitoring	Graph approaches to watching delegation and tool-call networks.
Multi-Agent Failure Analysis	Post-incident analysis patterns for multi-agent failures.
Runtime Telemetry Reference	Canonical telemetry fields, events, and formats.
Operational Metrics	Metrics catalogue for judges, guardrails, and the overall stack.

Control Catalogues

The working catalogues of controls, solutions, and hardening guidance.

Document	Description
Agentic Controls Catalogue	The working catalogue of controls for agentic deployments.
Agentic Controls Extended	Extended agentic control guidance and edge cases.
Technical Controls	Network, WAF, DLP, gateway controls for AI traffic.
Current Solutions	Industry solutions implementing this pattern.
Control Selection Guide	Methodology for choosing and sequencing controls.
AI Endpoint Hardening	Hardening the model endpoint, gateway, and tool server.
RAG Security	RAG pipeline security, from ingestion to retrieval.

Economics and Identity

Cost, governance, and non-human identity considerations.

Document	Description
Cost and Latency	Cost and latency impact of each control layer.
Economic Governance	Budget, attribution, and economic guardrails for AI workloads.
NHI Lifecycle	Non-human identity lifecycle management.
Supply Chain Controls	AI supply chain security controls, from weights to packages.
Humans in the Business Process	Where humans sit in the operational loop, and what they do there.