UNPKG

aiwg

Version:

Cognitive architecture for AI-augmented software development with structured memory, ensemble validation, and closed-loop correction. FAIR-aligned artifacts, 84% cost reduction via human-in-the-loop, standards adopted by 100+ organizations.

278 lines (193 loc) 7.01 kB
# Getting Started with Intake Intake is how AI agents learn about your project. It generates structured documents that guide all future operations - from architecture decisions to test strategies to deployment plans. These same documents serve your human team too: stakeholder alignment, onboarding new developers, planning sessions, and project documentation. One intake process, dual purpose. Just describe what you're building and let the system handle the rest. --- ## New Project **Natural language:** ```text "Start a new project for a payment processing API" ``` **Slash command:** ```text /intake-wizard "payment processing API" ``` The system: 1. Asks clarifying questions (--interactive by default) 2. Generates intake documents 3. Validates completeness 4. Kicks off inception phase 5. Assigns agents and begins work --- ## Existing Codebase **Natural language:** ```text "Analyze this codebase and start a project from it" ``` **Slash command:** ```text /intake-from-codebase ./src ``` The system: 1. Scans your code for patterns, dependencies, architecture 2. Generates intake reflecting what exists 3. Identifies gaps and technical debt 4. Kicks off with full context of current state --- ## Add Context with --guidance Shape the entire project with upfront context. **Natural language:** ```text "Start a new inventory management system" --guidance "must integrate with SAP, team is new to microservices" ``` **Slash command:** ```text /intake-wizard "inventory management system" --guidance "must integrate with SAP, team is new to microservices" ``` The guidance flows through everything: - Intake questions focus on SAP integration points - Architecture considers team experience level - Risk assessment flags microservices learning curve - All generated artifacts reflect these constraints ### More Examples ```text "Analyze this codebase" --guidance "preparing for SOC2 audit" /intake-from-codebase . --guidance "preparing for SOC2 audit" Focuses on security posture, access controls, logging "Start a mobile app project" --guidance "offline-first, rural users with poor connectivity" /intake-wizard "mobile app" --guidance "offline-first, rural users with poor connectivity" Prioritizes sync strategies, local storage, graceful degradation ``` --- ## Interactive Mode Ask for maximum collaboration with `--interactive`. **Natural language:** ```text "Start intake for a healthcare app" --interactive ``` **Slash command:** ```text /intake-wizard "healthcare app" --interactive ``` **Example conversation:** ```text Agent: Healthcare apps often have compliance requirements. Which apply? 1. HIPAA (US patient data) 2. GDPR (EU users) 3. Both 4. Not sure yet - help me figure it out > 4 Agent: Let's work through this. Will the app store any of these? - Patient names or contact info - Medical diagnoses or conditions - Prescription information > Yes, patient names and prescriptions Agent: That's Protected Health Information under HIPAA. I'll include: - HIPAA compliance requirements in intake - Security controls for PHI - Audit logging requirements - BAA considerations for third-party services ``` --- ## Combine --guidance and --interactive The most powerful pattern: **Natural language:** ```text "Start a fintech project" --interactive --guidance "Series A startup, need to move fast but can't compromise on security" ``` **Slash command:** ```text /intake-wizard "fintech project" --interactive --guidance "Series A startup, move fast, security non-negotiable" ``` Agent understands the context and asks smart questions: - "PCI compliance needed, or just general financial data?" - "In-house security team, or need automated tooling?" - "MVP scope - what's the minimum for launch?" --- ## Resume or Complete Intake **Natural language:** ```text "Continue with the intake we started" ``` **Slash command:** ```text /intake-wizard --complete ``` --- ## Start from Existing Intake Files If you already have intake documents in `.aiwg/intake/`: **Natural language:** ```text "Start the project from the intake files" ``` **Slash command:** ```text /intake-start .aiwg/intake ``` With guidance: ```text /intake-start .aiwg/intake --guidance "prioritize security requirements" ``` --- ## Marketing Projects For marketing campaigns instead of software: **Natural language:** ```text "Start a marketing campaign for our product launch" ``` **Slash command:** ```text /marketing-intake-wizard "Product launch campaign for Q1" ``` From existing materials: ```text /intake-from-campaign ./marketing-assets --interactive ``` --- ## What Happens Behind the Scenes ``` Your description ┌─────────────────────────────┐ Intake Wizard - Parses your description - Identifies project type - Asks clarifying questions│ - Generates intake docs └──────────────┬──────────────┘ ┌─────────────────────────────┐ Validation - Checks completeness - Identifies gaps - Suggests clarifications └──────────────┬──────────────┘ ┌─────────────────────────────┐ Inception Kickoff - Assigns specialist agents│ - Creates phase plan - Begins orchestrated work └─────────────────────────────┘ ``` --- ## Quick Reference | Action | Natural Language | Slash Command | |--------|------------------|---------------| | New project | "Start a new project for X" | `/intake-wizard "X"` | | From codebase | "Analyze this codebase" | `/intake-from-codebase .` | | With guidance | + `--guidance "context"` | + `--guidance "context"` | | Interactive | + `--interactive` | + `--interactive` | | Resume | "Continue intake" | `/intake-wizard --complete` | | From files | "Start from intake files" | `/intake-start .aiwg/intake` | | Marketing | "Start marketing campaign" | `/marketing-intake-wizard "X"` | --- ## Tips for Best Results 1. **Be specific about the domain** - "payment API" vs "payment API for recurring SaaS subscriptions" 2. **Include key constraints upfront** - Compliance, timeline, team size, integrations 3. **Use --guidance for context that affects everything** - Not just features, but business context 4. **Let --interactive ask questions** - The AI often identifies things you haven't thought about 5. **Don't worry about format** - Natural language works. The system figures out what you need.