The orchestration and security layer of the SAKE framework. It handles compiling, auditing, and dispatching tasks, as well as managing lineage, diffs, and patches.

An AI agent responsible for SAKE synthesis, converting TaskIR into .sake files and handling compilation.

Citadel Automation & Policy System. A grading system for risk and capability assessment of automated tasks.

Framework for Agent Trust and Evaluation. A system for scoring agent performance and trustworthiness.

A unique identifier assigned to every build, task, and artifact, providing an immutable audit trail.

Master Control Program. The central governance and tooling layer, providing a secure workspace and enforcing development policies.

Multidisciplinary Analysis and Optimization. A discipline for scoring simulations across cost, risk, and confidence dimensions.

The AI auditor agent, responsible for tracking lineage, gathering evidence, and enforcing governance proofs.

An AI agent that acts as an evaluator, performing quality and policy checks and scoring tasks against rubrics.

Secure, Auditable, Kinesthetic, Evolving. The core autonomous software development and maintenance framework within Citadel Nexus.

The governed lineage datastore where all audit trails, metrics, and evidence are stored.

Task Intermediate Representation. A standardized specification for tasks that flow through the SAKE system, which are then translated into executable actions.

An ecosystem for synthesizing, auditing, and housing knowledge, turning raw logs into validated information for other agents.

The primary conversational AI agent that acts as a guide and router, translating natural language into TaskIR.