DLC Overview

The AI-Driven Development Lifecycle (DLC) is Corpay’s methodology for building software with AI assistance while maintaining engineering discipline, governance, and auditability. CAT implements the DLC as structured workflows that guide you through four sequential phases.

Skills, Agents, and Personas

Three terms appear throughout CAT — here’s what they mean:

Term	Meaning
Skill	A markdown instruction set that tells the AI how to behave and what to produce for a specific workflow (e.g., cat-arc-builder)
Agent	Your AI assistant while it’s actively using a skill — it reads the skill file and operates within its rules
Persona	The named character the agent adopts during a session (e.g., Julian the Solution Architect). Personas give agents domain expertise, a communication style, and a defined role

A skill activates an agent; the agent adopts the persona. When you invoke cat-arc-builder, your AI becomes Julian — until the session ends or you invoke a different skill.

---

Why the DLC Exists

AI coding assistants can generate code fast — but speed without structure leads to:

• Code that doesn’t match business intent
• Architecture decisions made implicitly by AI, not explicitly by engineers
• No audit trail connecting requirements to implementation
• Compliance gaps that surface late in the process

The DLC solves this by providing a phase-gated lifecycle where AI assists at every step but humans remain in control of decisions.

The Four Phases

graph LR
    I[1. Intent] --> A[2. Architecture]
    A --> C[3. Construction]
    C --> V[4. Validation]

    style I fill:#E3F2FD,stroke:#1565C0
    style A fill:#E8F5E9,stroke:#2E7D32
    style C fill:#FFF3E0,stroke:#E65100
    style V fill:#F3E5F5,stroke:#6A1B9A

Phase 1: Intent — “What and Why”

Capture what you’re building and why it matters. Every initiative starts here.

Key artifacts: Intent Statement, Intent Brief, Feature List Key agent: Clara (Product Intent Manager) Key question: “Can the Solution Architect start the ARC without calling a meeting?”

If the answer is yes, your intent is clear enough.

→ Intent Phase Deep Dive

Phase 2: Architecture — “How, with Constraints”

Define how to build it, expressed as enforceable constraints. The ARC (Architecture Reference Contract) is not a suggestion — it’s a contract that governs all construction.

Key artifacts: ARC, AI Validation Report, Readiness Report Key agent: Julian (Solution Architect) Key question: “Is every constraint traceable to a business need?”

→ Architecture Phase Deep Dive

Phase 3: Construction — “Build It, Bolt by Bolt”

Implement the solution in discrete, well-scoped bolts. Each bolt maps to specific ARC constraints and has explicit acceptance criteria.

Key artifacts: Bolt Specs, Code, Code Reviews, Test Plans Key agents: Marcus (Developer), Vera (QA) Key question: “Does this bolt satisfy every ARC constraint it maps to?”

→ Construction Phase Deep Dive

Phase 4: Validation — “Did We Build the Right Thing?”

Verify that the implementation matches intent, satisfies architecture constraints, and meets compliance requirements.

Key artifacts: ARC Adherence Report, Work Acceptance, Compliance Report Key agents: Vera (QA), Clara (PI) Key question: “Can we ship with confidence?”

→ Validation Phase Deep Dive

Phase Gates

Each phase has a gate — a checkpoint that validates readiness before proceeding:

Transition	Gate	Workflow
Intent → Architecture	Intent Checkpoint	cat-intent-checkpoint
Architecture → Construction	ARC Checkpoint + AI Validation + Readiness Check	cat-arc-checkpoint, cat-ai-validation, cat-readiness-check
Construction → Validation	Code Review	cat-code-review
Validation → Done	Compliance Report	cat-compliance-report

Gates are recommended, not mandatory. You can proceed without running a gate workflow, but doing so leaves a compliance gap — the cat-compliance-report will flag it as missing evidence. Gates are most valuable before irreversible decisions: starting Architecture before Intent is clear adds rework risk; starting Construction before the ARC is validated risks misaligned implementation. For low-risk brownfield changes, the Quick Track’s reduced gate set is explicitly supported.

Artifacts Flow

Every phase produces artifacts that feed the next:

Intent Statement → Intent Brief → Feature List
                                        ↓
                                       ARC
                                        ↓
                              Bolt Specs (per feature)
                                        ↓
                              Implementation + Tests
                                        ↓
                    ARC Adherence → Work Acceptance → Compliance Report

All artifacts are markdown files stored in _cat/artifacts/ with full traceability.

Anytime Workflows

Some workflows aren’t tied to a specific phase — use them at any point:

Workflow	Purpose
cat-help	Context-aware guidance — what to do next
cat-dlc-status	Artifact completion dashboard
cat-party-mode	Multi-agent roundtable discussions
cat-advanced-elicitation	Deep critical thinking on any artifact
cat-meeting-digest	Convert meeting transcripts to DLC-tagged notes
cat-document-project	Generate project documentation

How CAT Manages Context

CAT workflows are powered by AI language models with limited context windows. The DLC’s artifact chain solves this by giving each workflow a specific, bounded reading list — workflows read targeted upstream artifacts rather than requiring the AI to hold the entire project in memory at once. This also creates a natural audit trail: each artifact documents its inputs, so the reasoning behind every decision is traceable.

Workflow	Reads
cat-intent-brief	Intent Statement
cat-arc-builder	Intent Brief, Feature List
cat-bolt-writer	ARC, Feature List
cat-bolt-execution	Bolt Spec, ARC
cat-arc-adherence-check	ARC, Bolt Execution reports
cat-compliance-report	All artifacts

Project Context: Run cat-generate-project-context early to create a project-context.md file that teaches all workflows your codebase’s conventions, stack, and patterns. This is especially valuable for brownfield projects.

Best practices:

1. Keep artifacts in the standard _cat/artifacts/ locations — workflows depend on these paths.
2. Reference specific artifacts by name (e.g., “build an ARC based on the password-reset Intent Brief”) to help workflows find the right context.
3. Use cat-document-project before starting DLC work on existing codebases.

---

See Also: Workflow Map · Getting Started · Glossary