War Discovery Interview

Discovery Interview — AWS Well-Architected Review

Conduct a structured interview to capture solution requirements and design intent that cannot be derived from code or documentation alone. Produces DESIGN_REQUIREMENTS.md as input for the discovery analyst agent.

Rules

• Structured, not open-ended. Each round has specific questions. Do not free-form explore.
• One round at a time. Present 2-4 questions per round via AskUserQuestion. Wait for answers before proceeding.
• Reference existing deliverables. Read prior phase outputs first. Never ask about something already documented.
• Respect "not applicable." If the user says a domain does not apply, record that and move on.
• No recommendations. Capture requirements, do not evaluate them. Analysis belongs to the discovery analyst.
• Write once at the end. Accumulate answers across all rounds, then write the output document.

Prerequisites

The orchestrator provides the output directory path. Expect these deliverables to exist from prior phases:

• DOCUMENT_CATALOGUE.md — Phase 1 output
• CODE_CATALOGUE.md — Phase 1 output
• DOCUMENT_ARCHITECTURE_REVIEW.md — Phase 2 output
• CODE_ARCHITECTURE_REVIEW.md — Phase 2 output

If any deliverable is missing, note its absence but proceed with the interview. The discovery analyst will flag the gap.

Step 1 — Read Prior Deliverables

Read all 4 deliverables from the output directory. Extract:

• What the system does (from catalogues)
• Architectural patterns and technology choices (from reviews)
• Known gaps and concerns already identified (from review gap sections)

Use this context to skip questions the deliverables already answer and to ask sharper follow-up questions where the deliverables are silent or ambiguous.

Step 2 — Business Objectives (Round 1)

Use AskUserQuestion to gather:

• What is the primary business objective this workload serves?
• Who are the end users and what are their expectations (availability, latency, geographic distribution)?
• What are the consequences of this workload being unavailable for 1 hour? 1 day?
• Are there regulatory or contractual SLAs that apply?

Step 3 — Workload Characteristics (Round 2)

Use AskUserQuestion to gather:

• What is the expected traffic pattern? (steady, spiky, seasonal, event-driven)
• What is the expected scale? (requests/sec, data volume, concurrent users)
• What are the data retention requirements?
• Is this workload stateless or stateful? Where does state live?

Step 4 — Compliance and Governance (Round 3)

Use AskUserQuestion to gather:

• Which compliance frameworks apply? (SOC 2, HIPAA, PCI-DSS, FedRAMP, GDPR, none, other)
• Are there data residency requirements? (specific AWS regions or countries)
• What is the data classification? (public, internal, confidential, restricted)
• Are there audit logging or evidence retention requirements?

Step 5 — Security Assessment (Round 4)

Use AskUserQuestion to gather:

• How do end users authenticate? (Cognito, SAML/SSO, API keys, IAM, other)
• How are secrets and credentials managed? (Secrets Manager, Parameter Store, vault, environment variables)
• Are there network isolation requirements? (VPC, private subnets, VPN, PrivateLink)
• What is the encryption posture? (at-rest, in-transit, KMS managed keys, default)

Step 6 — Performance Requirements (Round 5)

Use AskUserQuestion to gather:

• What are the latency targets for critical operations? (p50, p99, or general expectations)
• Are there batch processing or async workloads? What are their SLAs?
• What caching strategy is in place or planned?
• Are there known performance bottlenecks or concerns?

Step 7 — Cost Analysis (Round 6)

Use AskUserQuestion to gather:

• Is there a monthly or annual budget target for this workload?
• What cost optimization strategies are in place? (reserved instances, savings plans, spot, right-sizing)
• Are there cost allocation or chargeback requirements? (tags, accounts, cost centers)
• Which cost dimensions matter most? (compute, storage, data transfer, third-party services)

Step 8 — Write DESIGN_REQUIREMENTS.md

Compile all answers into a structured document. Write to the output directory path provided by the orchestrator.

# Design Requirements

## Business Objectives

[Compiled from Round 1 answers. Include business objective, end user expectations, downtime impact, SLAs.]

## Workload Characteristics

[Compiled from Round 2 answers. Include traffic pattern, scale, data retention, state management.]

## Compliance and Governance

[Compiled from Round 3 answers. Include frameworks, data residency, classification, audit requirements.]

## Security Assessment

[Compiled from Round 4 answers. Include authentication, secrets management, network isolation, encryption.]

## Performance Requirements

[Compiled from Round 5 answers. Include latency targets, async workloads, caching, known bottlenecks.]

## Cost Analysis

[Compiled from Round 6 answers. Include budget, optimization strategies, allocation, key cost dimensions.]

## Context from Prior Phases

[Brief summary of what the catalogues and architecture reviews already established. Note any areas where interview answers confirm, contradict, or extend the documented architecture.]

Write the file, then confirm to the user that DESIGN_REQUIREMENTS.md has been written and the interview is complete.