natural-language-sdlc-processor

name	Natural Language SDLC Processor
description	Understands natural English language queries and converts them to SDLC commands. Supports hybrid pattern matching + AI understanding. Asks for ADO work item if missing, shows non-closed items for quick selection. Works with any stage, any role.
tags	["ai-sdlc","natural-language","workflow-orchestration"]
triggerPhrases	["create a master story for","review this PRD for","generate tech tasks for","check gates for","validate release for","run the full workflow","what's the status of","I need a","can you help me with","let me do the whole pipeline"]

Natural Language SDLC Processor

Converts natural English language queries into SDLC slash commands and executes them with interactive prompts.

How It Works

1. Input Processing

When the user types natural language (e.g., "Create a master story for multi-language support"):

1. Pattern Matching (fast):

   • Detect common phrases: "create story" → /project:grooming --type=master
   • Detect actions: "review", "generate", "check", "validate"
   • Map to SDLC commands

2. AI Understanding (fallback):

   • If no pattern match, Claude interprets intent
   • Suggests best command + parameters
   • Offers options if ambiguous

2. Parameter Resolution

If user doesn't provide ADO work item number (AB#123):

1. Ask via text: "What work item? (e.g., AB#123)"
2. Show recent items: Load all non-closed items from user's ADO project
3. User selects: Pick from menu or type custom ID

3. Execution

• Route to appropriate command skill
• Pass through interactive prompts
• Execute with full context
• Save to memory

Natural Language Examples

User: "Create a master story for multi-language support"
→ System detects: "create" + "master story"
→ Matches: /project:grooming --type=master
→ Asks: "What's the work item? (e.g., AB#123)"
→ Asks: "Happy path only or comprehensive?"
→ Executes /project:grooming

User: "Review this PRD for gaps"
→ System detects: "review" + "PRD"
→ Matches: /project:prd-review
→ Asks: "Review from which perspective? (Product/Backend/Frontend/QA)"
→ Executes /project:prd-review

User: "Let me do the whole pipeline for this feature"
→ System detects: "whole pipeline" / "full workflow"
→ Triggers: Orchestrator Agent
→ Shows: Progress dashboard with 15-stage workflow
→ Executes: Complete workflow PRD → Production

User: "Check gates for this release"
→ System detects: "check gates"
→ Matches: /project:gate-check
→ Asks: "Which stage? (1-15)"
→ Executes: Gate validation

Command Pattern Mapping

User Intent	Pattern	Command	Example
"set up" / "install" / "configure"	setup/install/configure/init/bootstrap/get started	/project:setup	"Set up the SDLC plugin"
"create requirement"	new feature/requirement/capture	/project:requirement-intake	"I have a new feature idea"
"create master story"	create + story	/project:grooming --type=master	"Create a master story for auth system"
"break into sprints"	break + sprints/split	/project:grooming --type=sprint	"Break this into sprint stories"
"review PRD"	review + prd	/project:prd-review	"Review this PRD for gaps"
"architecture"	architecture/design review/ADR	/project:architecture-review	"Review architecture for this"
"system design"	system design/data model/API design	/project:system-design	"Design the data model"
"plan sprint"	sprint plan/task breakdown	/project:sprint-planning	"Plan sprint 12 tasks"
"generate tech tasks"	generate + tasks/implement	/project:implementation	"Generate tech tasks from this story"
"code review"	code review/PR review	/project:code-review	"Review this pull request"
"test plan"	test + plan/QA/test cases	/project:test-design	"Create a test plan for this"
"perf test"	performance/load test/NFR	/project:performance-testing	"Run load test for this API"
"staging"	staging/SIT/UAT	/project:staging-validation	"Deploy to staging for testing"
"release prep"	prepare/release/go-live prep	/project:release-prep	"Prepare this for release"
"deploy"	deploy + prod/production	/project:deployment	"Deploy to production"
"monitor"	monitor + setup/alerts/dashboard	/project:monitoring	"Set up monitoring for this service"
"incident"	incident/post-mortem	/project:incident-response	"Create post-mortem for this incident"
"full workflow"	workflow/pipeline/full SDLC	Orchestrator Agent	"Let me do the full workflow"

Smart Defaults

When information is missing, use context:

• Role: Detect from user context (if provided), else ask
• Stack: Check .sdlc/state.json for previous choice
• Stage: Infer from command type or ask
• Estimation: Default to "balanced" unless specified

Integration Points

• All 15 command skills: Route patterns to appropriate skill
• Orchestrator Agent: Handle multi-step workflows
• Memory system: Load previous context, save outcomes
• ADO: Fetch non-closed items for selection
• Context recovery: Use .sdlc/state.json for smart defaults

Error Handling

If pattern doesn't match:

1. Ask user to clarify
2. Show available commands
3. Or offer conversation mode
4. Never assume intent