Skip to content

Glossary

Quick reference for terms used throughout CAT and the DLC.


Acceptance Criteria — Specific, testable conditions that must be true for a bolt to be considered complete. Each criterion maps to at least one test case.

ADO — Azure DevOps. Microsoft’s project management and CI/CD platform. CAT can generate ADO Features and Stories from architecture documents via cat-arch-to-stories.

Advanced Elicitation — A technique for pushing the AI to reconsider and improve its output using methods like Socratic questioning, first principles, pre-mortem, or red-teaming. Invoked via cat-advanced-elicitation or the “A” collaboration menu option.

Agent — An AI persona with a specific domain of expertise, communication style, and set of principles. CAT has 9 agents: Janus, Livia, Clara, Flora, Julian, Marcus, Argus, Vera, and Tullia. See Agent Reference.

AI Governance — The section of every ARC that defines boundaries, review requirements, and testing expectations for AI-assisted development. A first-class DLC concern, not an afterthought.

AI Validation — A governance gate where the AI produces evidence that it correctly understands the Intent Brief and ARC before construction begins. Created by cat-ai-validation.

ARC — Architecture Reference Contract. The central governance document defining enforceable constraints (MUST, MUST NOT, SHOULD, MAY) that govern all construction. Not a suggestion — a contract. See What Is an ARC?.

ARC Checkpoint — A gate workflow (cat-arc-checkpoint) that validates the ARC for completeness and alignment with the Intent Brief. Runs after ARC creation and before AI Validation and Readiness Check, gating the transition to Construction.

ARC Adherence Check — Post-construction verification that every ARC constraint was honored in the implementation. Produces a constraint-by-constraint compliance matrix.

Artifact — A structured markdown document produced by a CAT workflow. All artifacts are stored in _cat/artifacts/ with subfolders for each phase.

Bolt — The atomic unit of construction in the DLC. A well-scoped piece of work with specific ARC constraint mappings, acceptance criteria, and testing requirements. See What Are Bolts?.

Brownfield — An existing codebase or system that is being modified or extended, as opposed to a greenfield (new from scratch) project. CAT provides Quick Track workflows optimized for brownfield changes.

Bolt Execution — The process of implementing a bolt against ARC constraints. Marcus enforces constraint compliance during implementation.

Bolt Spec — The specification document for a bolt, including scope, ARC constraint mappings, acceptance criteria, testing requirements, and dependencies.

CAT — Corpay AI-DLC Toolkit. The AI-assisted toolkit that implements the DLC through structured workflows, agent personas, and governance artifacts.

Change Proposal — A formal document for managing mid-construction scope changes. Includes old→new diff across Intent Brief, ARC, and affected bolts.

Collaboration Menu — The A/P/C menu presented after content-generating steps: A (Advanced Elicitation), P (Party Mode), C (Continue).

Compliance Controls — The 10 controls that make up the DLC compliance framework, evaluated by the Compliance Report. See Compliance Controls.

Compliance Report — The capstone validation artifact — a full 10-control compliance matrix covering all DLC phases. Created by cat-compliance-report.

Constraint — An enforceable rule in the ARC using RFC 2119 keywords (MUST, MUST NOT, SHOULD, MAY). Each constraint has a unique ID and traces to a business need.

DLC — AI-Driven Development Lifecycle. Corpay’s methodology for building software with AI assistance while maintaining engineering discipline, governance, and auditability.

Feature — A unit of work decomposed from the Intent Brief. Each feature gets its own set of bolts during Construction.

Frontmatter — YAML metadata at the top of a markdown file, delimited by ---, defining properties like title and description. CAT artifacts use frontmatter for compliance tracking and tool discovery.

Feature List — A prioritized list of features decomposed from the Intent Brief. Created by cat-features-create.

Full Flow — The complete DLC workflow: Intent Statement → Intent Brief → Features → ARC → Bolts → Validation. Used for larger initiatives.

Intent Brief — The operational expansion of an Intent Statement, adding scope, metrics, risks, NFRs, and dependencies. The primary input for ARC creation.

Intent Checkpoint — A validation gate that checks Intent artifacts for completeness, specificity, and compliance readiness before proceeding to Architecture.

Intent Statement — The starting point for any DLC work. Captures who, what problem, why now, expected outcome, and success metrics.

Module — A collection of related CAT workflows and agents. Current modules: cat-core (utilities), core (foundation), aidlc (DLC lifecycle).

NFR — Non-Functional Requirement. A requirement that defines how a system should operate (performance, security, reliability, scalability) rather than what it should do. NFRs are captured in the Intent Brief and enforced as ARC constraints.

Party Mode — A multi-agent roundtable discussion where relevant agent personas provide diverse perspectives on an artifact or decision. Invoked via cat-party-mode or the “P” collaboration menu option.

Phase — One of four sequential stages in the DLC: Intent, Architecture, Construction, Validation.

Phase Gate — A checkpoint that validates readiness before proceeding to the next phase (e.g., Intent Checkpoint, ARC Checkpoint, Readiness Check).

PRFAQ — Press Release / Frequently Asked Questions. Amazon’s “working backwards” technique where a product initiative begins by writing the launch press release and FAQ as if the product is already shipped. CAT provides cat-prfaq to facilitate this exercise.

Project Context — A project-context.md file that captures technology stack, coding patterns, and conventions. Referenced by all CAT workflows for project-specific guidance.

Quick ARC — A lightweight ARC for small brownfield changes, focusing on what’s changing rather than the full system architecture.

Quick Intent — A combined Intent Statement + Intent Brief for small changes (1–3 bolts). Faster than the full two-step flow.

Quick Track — The streamlined DLC path using Quick Intent and Quick ARC for small, well-understood changes. See Choosing Your Track.

Readiness Check — A pre-construction traceability audit that verifies connections across Intent Brief → ARC → Features → Bolts.

Skill — A specific capability or workflow implemented as a SKILL.md file in the _cat/ folder. Skills are invoked as commands in the AI IDE.

Traceability — The ability to follow a thread from a business need (Intent) through architecture constraints (ARC) to specific code (Bolts) and verification (Validation).

Workflow — A structured, AI-facilitated process that produces one or more artifacts. Each workflow has a specific agent, input requirements, and outputs.


See Also: DLC Overview · Getting Started · Workflow Map