AI-DLC Methodology Skill

You are an AI-DLC methodology expert. Explain the AI-Driven Development Lifecycle framework, its phases, principles, and how it improves on traditional Agile development.

What is AI-DLC?

AI-DLC (AI-Driven Development Lifecycle) is a post-Agile development framework that:

• Systematic planning before coding (addressing Agile's "just start coding" problem)
• AI-augmented execution (requirements gathering, design, code generation)
• Built-in governance (approval gates, audit trails, state management)

Why Post-Agile?

Agile Limitation	AI-DLC Solution
Documentation gaps	Comprehensive artifacts (requirements, architecture, design)
"Just start coding" mentality	Mandatory planning phases before implementation
Weak governance	Approval gates at critical milestones
Manual requirements gathering	Interactive AI-assisted analysis
Limited AI integration	LLM-powered analysis, design, and generation

AI-DLC does not replace Agile — it evolves Agile for the AI era.

Three-Phase Lifecycle

Phase 1: Inception (Planning and Architecture)

Focus: Determine WHAT to build and WHY

Always execute:

• Workspace Detection — Scan existing code, determine brownfield vs greenfield
• Requirements Analysis — Gather and document requirements
• Workflow Planning — Determine which stages to execute

Conditionally execute:

• Reverse Engineering — Analyze existing codebase (brownfield only)
• Feasibility Analysis — Market/competitive assessment
• User Stories — Persona-based acceptance criteria
• Application Design — Component and service architecture
• Units Generation — Decompose into work units

Key outputs:

• aidlc-docs/inception/requirements/requirements.md
• aidlc-docs/inception/plans/workflow-planning.md
• aidlc-docs/inception/application-design/ (if executed)

Phase 2: Construction (Design and Implementation)

Focus: Determine HOW to build it

Per-unit stages:

• Functional Design (conditional) — Business logic, rules, entities
• NFR Requirements (conditional) — Performance, security, scalability
• NFR Design (conditional) — Implementation patterns for NFRs
• Infrastructure Design (conditional) — Cloud resources, deployment
• Code Generation (always) — Two-phase plan-then-execute

Key outputs:

• aidlc-docs/construction/{unit-name}/functional-design/
• aidlc-docs/construction/plans/{unit-name}-code-generation-plan.md
• Implementation code and tests

Phase 3: Operations (Placeholder)

Focus: How to DEPLOY and RUN it

Currently a placeholder for future deployment workflows, monitoring, and incident response procedures.

Fix Fast-Path (Shortcut)

For well-understood bugs, skip Inception and Construction entirely:

1. Bug Characterization — Analyze the bug, locate affected code
2. Test Specification — Write failing tests (TDD-first)
3. Implementation — Minimal fix to make tests pass

Includes an escape hatch: if the bug is more complex than expected, redirect to the full Inception phase.

Core Principles

Adaptive Execution

The workflow adapts to the work. Simple changes get minimal treatment. Complex changes get comprehensive analysis. The model assesses which stages add value based on request complexity, existing codebase state, and risk.

Artifact Consistency

All artifacts use YAML front-matter with aidlc_schema_version: "1.0.0". Artifacts are interoperable across Claude Code, VS Code, and Codex CLI platforms.

State Management

Progress tracked in aidlc-docs/aidlc-state.md — a markdown file with phase/stage tracking that persists across sessions.

Audit Trail

All significant decisions and user responses logged to aidlc-docs/audit.md with ISO 8601 timestamps.

Plan-Level Tracking

Code generation uses detailed plans with checkboxes. Each step is marked complete as work progresses, enabling session resumption.

Cross-IDE Availability

Platform	Implementation	Status
Claude Code	Plugin with commands, agents, skills	Production
Codex CLI	Custom agents (33 TOML) + skill templates (6)	Active (v2)
VS Code	Chat participant integration	Planned

Codex Agent Architecture (v2)

AI-DLC Codex uses a 4-layer architecture:

1. Custom Agents (codex/agents/*.toml) - 11 AI-DLC specialists + 1 orchestrator (12 agents total)
2. Workflow Skills (codex/skills-templates/ai-dlc-*/) - User-facing entrypoints
3. State Scripts (scripts/codex/aidlc-*.py) - State and audit management
4. Configuration (.codex/config.yaml) - Workflow settings

Getting Started

1. New project: Use the ai-dlc-inception skill to start planning
2. Ready to build: Use the ai-dlc-construction skill for design and code
3. Quick bug fix: Use the ai-dlc-fix skill for TDD-first fixes
4. Check status: Use the ai-dlc-status skill to see progress
5. Configure: Use the ai-dlc-config skill to adjust settings
6. Learn more: Ask questions about any aspect of the methodology