AI-DLC Help

Explain the AI-DLC methodology and guide users through the planning plugin.

References

• •Use @${CLAUDE_PLUGIN_ROOT}/references/aidlc-methodology.md for detailed methodology documentation
• •Use @${CLAUDE_PLUGIN_ROOT}/references/planning-shared.md for templates and operational guidance

What is AI-DLC?

AI-DLC (AI-Driven Development Lifecycle) is a methodology that puts AI at the center of the development process. It was developed by AWS as an AI-native approach to software engineering.

Core Idea

Traditional methods: Human initiates, AI assists AI-DLC: AI proposes, human approves

This reversal allows developers to focus on high-value decision-making while AI handles planning, task decomposition, and execution.

Key Principles

1. •Reimagine Rather Than Retrofit - Don't force AI into old methods; design for AI capabilities
2. •Reverse the Conversation Direction - AI drives workflows, humans validate and approve
3. •Integration of Design Techniques - DDD, BDD, TDD are core, not optional add-ons
4. •Align with AI Capability - Balance AI strengths with human oversight
5. •Build Complex Systems - Designed for architectural complexity, not simple scripts
6. •Retain Human Symbiosis - Keep artifacts that enable validation and risk mitigation
7. •Facilitate Transition - Familiar concepts with modernized terminology
8. •Streamline Responsibilities - Developers transcend traditional silos
9. •Minimize Stages, Maximize Flow - Continuous iteration with strategic checkpoints
10. •No Opinionated Workflows - AI recommends approach based on context

Core Artifacts

Available Skills

Workflow Order

1. /aidlc-intent
   |
   v (Intent approved in Confluence)
2. /aidlc-elaborate
   |
   v (Tasks & Units created in Confluence, Bolt groupings proposed, reviewed, reorganized)
3. /aidlc-design
   |
   v (Domain model & ADRs documented, confidence assessed)
4. /aidlc-verify
   |
   v (Documentation verified, Bolts refined, transferred to Jira)
5. /aidlc-bolt
   |
   v (Bolt implementation with TDD)

Phase 1: Intent Documentation (/aidlc-intent)

Creates an Intent document in Confluence containing:

• •Problem/Opportunity statement
• •Target users and outcomes
• •Scope (in/out)
• •Technical considerations
• •NFRs and measurement criteria
• •Risks and assumptions
• •Proposed Units (hypotheses)

Output: Approved Confluence document with workflow status tracking

Phase 2: Decomposition (/aidlc-elaborate)

Breaks the Intent into actionable work using Mob Elaboration:

1. •Theme clusters identified
2. •Tasks elaborated in parallel (subagents)
3. •Tasks grouped into Units
4. •Bolt groupings proposed for each Unit
5. •Confluence pages created for review
6. •Team reviews and comments
7. •Comments resolved, Tasks refined
8. •Units reorganized based on domain principles

Output: Confluence pages for Units and Tasks with proposed Bolt groupings (Jira transfer happens later in verify phase)

Phase 3: Design (/aidlc-design)

Creates design artifacts for implementation with confidence assessment:

1. •Assess context sufficiency (confidence check)
2. •Domain models (aggregates, entities, value objects)
3. •Logical design (patterns, NFR solutions)
4. •Architecture Decision Records (ADRs)

Output: Design documentation linked to Units in Confluence

Phase 4: Verification (/aidlc-verify)

Verifies documentation completeness and transfers to Jira:

1. •Spawn parallel sub-agents to assess each Unit
2. •Calculate confidence score across all documentation
3. •Identify gaps and provide remediation suggestions
4. •Refine Bolt groupings based on assessment
5. •If confidence ≥80%, transfer to Jira:
   • •Units → Sub-epics
   • •Bolts → Stories
   • •Tasks → Sub-tasks
6. •Clean up Confluence decomposition pages

Output: Jira artifacts (Sub-epics → Stories → Sub-tasks) linked to Intent

Quick Start Guide

Starting Fresh?

"I want to plan a new feature for user authentication" Use /aidlc-intent

This will:

1. •Gather requirements through clarifying questions
2. •Draft an Intent document
3. •Create the Confluence page for team review

Have an Approved Intent?

"Break down the authentication intent into Tasks" Use /aidlc-elaborate

This will:

1. •Validate the Intent is approved
2. •Identify theme clusters
3. •Spawn parallel agents to elaborate Tasks
4. •Create Confluence pages for Tasks and Units
5. •Propose Bolt groupings for each Unit
6. •Guide you through review and reorganization

Note: Jira transfer happens later in /aidlc-verify after design is complete.

Ready to Design?

"Create the domain model for the auth unit" Use /aidlc-design

This will:

1. •Validate Units exist in Confluence
2. •Assess context sufficiency (confidence check)
3. •Create domain models using DDD principles
4. •Document logical design decisions
5. •Create ADRs for architectural choices

Ready to Transfer to Jira?

"Verify documentation and transfer to Jira" Use /aidlc-verify

This will:

1. •Spawn sub-agents to assess each Unit's documentation
2. •Calculate confidence score (needs ≥80% to proceed)
3. •Identify gaps and suggest fixes
4. •Refine Bolt groupings for each Unit
5. •Transfer to Jira: Units → Sub-epics, Bolts → Stories, Tasks → Sub-tasks
6. •Clean up Confluence decomposition pages

Key Concepts Explained

Intent vs Epic

Unit vs Sub-epic

Task vs Sub-task

Bolt vs Sprint

Mob Elaboration

A collaborative ritual where AI and humans work together:

• •Single room (physical or virtual) with shared screen
• •AI proposes breakdown of Intent into Tasks, Units, and Bolt groupings
• •Team reviews, challenges, and refines
• •Condenses weeks of work into hours

Response Behavior

When this skill is invoked:

1. •Greet the user and acknowledge their question
2. •Determine their need:
   • •General methodology questions -> Reference aidlc-methodology.md
   • •Specific skill usage -> Provide targeted guidance
   • •Workflow questions -> Explain the process flow
   • •Getting started -> Suggest the appropriate skill
3. •Provide clear, concise guidance
4. •Suggest next steps based on their context

Example Interactions

User: "What is AI-DLC?" Response: Explain the core concept, key principles, and how it differs from traditional methods.

User: "How do I start planning a new feature?" Response: Recommend /aidlc-intent, explain what it does, and what they'll need (project context, stakeholder info).

User: "What's the difference between a Unit and an Epic?" Response: Explain that Units are cohesive work elements from AI-DLC, while Epics are Jira's work tracking. Units become Sub-epics in Jira.

User: "Explain Mob Elaboration" Response: Describe the collaborative ritual, its participants, AI's role, and the outputs.

Troubleshooting

"I don't have a Confluence doc yet"

Start with /aidlc-intent to create the Level 1 Intent document.

"I have a Confluence doc but it's not approved"

Review the Workflow Status table in the doc. If "Intent" is not "Approved", gather stakeholder approval before proceeding to decomposition.

"I want to skip the Confluence phase"

While possible with explicit override, Confluence-first is recommended for:

• •Team collaboration and review
• •Comment resolution before Jira creation
• •Traceability between artifacts

"The skill said my prerequisites are incomplete"

Check that prior phases are complete:

• •For /aidlc-elaborate: Need approved Intent in Confluence
• •For /aidlc-design: Need Units created in Confluence (from elaborate phase)
• •For /aidlc-verify: Need design documentation and Bolt groupings complete

Further Reading

For detailed methodology documentation, ask about specific topics:

• •"Tell me about the 10 key principles"
• •"Explain the Construction Phase"
• •"What are Domain Design artifacts?"
• •"How does AI-DLC handle brown-field development?"

The methodology reference contains the complete AWS AI-DLC method definition with examples and prompts.