Core responsibilities
- Own the technical architecture for one or more domains
- Review and approve all design decisions before implementation
- Define data models and integration contracts across systems
- Mentor engineers on architecture patterns and trade-offs
- Drive technical due diligence on new initiatives
Where AI saves you the most time
- Generating Mermaid sequence and architecture diagrams from code or specs
- First-pass code review across every PR via Copilot Coding Agent
- Optimizing SQL queries and validating schemas in plain English
- Drafting ADRs (Architecture Decision Records) from meeting notes
- Producing API documentation from working code
Key AI tools for this role
- GitHub Copilot — primary diagram and documentation generator
- Copilot Coding Agent — automated code review
- Mermaid Preview — live diagram visualization
- Oracle MCP — natural-language database analysis
.prompt.md files in .github/prompts/.
generate-architecture.prompt.md
# generate-architecture.prompt.md --- title: Architecture Diagram Generation (Simple) description: Generates a high-level Mermaid architecture diagram focused on a specific functional area. --- # Role You are a senior solution architect. # Task Read the following requirement document and produce a high-level system architecture diagram in Mermaid. Focus only on [FUNCTIONAL_AREA] functional requirement. Keep the architecture simple. # Output A single Mermaid diagram saved to architecture/[feature]_architecture.mmd # Constraints Use clear, unambiguous labels. Keep the diagram minimal but complete — show all major components and their relationships. Do not include implementation details — focus on system boundaries and data flow.
generate-architecture-detailed.prompt.md
# generate-architecture-detailed.prompt.md --- title: Architecture Diagram Generation (Detailed) description: "Full component architecture with data flows, protocols, auth, and deployment environment." --- # Role Senior Solution Architect # Task Produce a detailed system architecture diagram in Mermaid from the provided requirements. # Diagram Must Include **Components:** Frontend, backend services, data stores, external systems, queues, APIs, IAM **Data Flow:** Integration points and system boundaries (client / service / data / external) **Protocols:** HTTP/REST/gRPC, auth (OAuth2/JWT), deployment environment (Azure/AWS) **Technology:** Keep technology-agnostic unless explicitly stated in the requirements # Output A Mermaid diagram. Use clear labels. Avoid ambiguous names. Keep minimal but complete — every major component and its connections must be shown.
sql-optimize.prompt.md
# sql-optimize.prompt.md --- title: SQL Query Optimization description: Analyzes a slow SQL query for performance issues and generates an optimized version with index recommendations. --- # Role You are a senior database architect and SQL performance specialist. # Task Analyze the following SQL query and provide: 1. **Execution Plan Analysis** Show an estimated execution plan and highlight performance bottlenecks. 2. **Index Recommendations** Identify missing indexes and provide the CREATE INDEX statements needed. 3. **Optimized Query** Rewrite the query for better performance. Add comments explaining non-trivial parts (window frames, GROUPING SETS, CTEs). 4. **Risk Check (if UPDATE/DELETE)** Generate a SELECT preview of affected rows before the destructive action. # Schema Context [Paste one-line schema hint; e.g. "Table dbo.Orders(order_id PK, customer_id, order_date, total_amount)"] # SQL Dialect [Specify: Oracle | SQL Server | PostgreSQL] # Query to Optimize [PASTE_QUERY_HERE]
GitHub Copilot
Architecture & Code AI
Your primary tool for generating Mermaid architecture diagrams, optimizing SQL queries, producing API documentation, and drafting technical specifications from requirements.
When to use
Architecture design, documentation, and technical analysis tasks — daily.
Copilot Coding Agent
Automated Code Review
Assign Copilot as a reviewer on any GitHub pull request. It reviews all changed files, produces file-level summaries, and leaves inline comments — replacing or augmenting your manual review process.
When to use
On every pull request — as a first-pass reviewer before human review.
Mermaid Preview
Diagram Visualization
VSCode extension that renders Mermaid diagrams generated by Copilot in a live preview panel. Essential for visualizing and iterating on architecture diagrams without leaving VSCode.
When to use
Any time Copilot generates a Mermaid diagram — preview it immediately via CTRL+SHIFT+P → Mermaid Preview.
Oracle MCP
Database Analysis
Query Oracle databases in plain English to analyze table sizes, compare environments, detect schema drift, and validate data — essential for data architecture validation tasks.
When to use
Data architecture validation, schema analysis, and cross-environment comparisons.
Getting set up — 6 steps
Request GitHub Copilot Access
Submit an access request through the Cisco AppStore. Install VSCode 1.109+ from the AppStore.
Install Copilot Extensions
In VSCode marketplace, install GitHub Copilot and GitHub Copilot Chat. Authenticate with your GitHub account.
Install Mermaid Preview
In VSCode marketplace, search for Mermaid Preview by Mermaid OSS. Install and confirm it appears in your extensions. Use CTRL+SHIFT+P → "Mermaid Preview: Preview Diagram" to activate.
Set Up Oracle MCP (Optional)
Install the Oracle SQL Developer VSCode extension. Connect to Cisco VPN. Contact [email protected] for test instance details. Add Oracle MCP to mcp.json.
Enable Copilot Coding Agent on GitHub
On any GitHub PR, go to Reviewers → search for Copilot → select it as a reviewer. The agent will analyze the PR and post comments automatically.
Run Your First Architecture Generation
Drag a requirements document into Copilot Chat, use the architecture prompt template, and select Claude Sonnet 4.5. Preview the Mermaid output immediately. Iterate until it matches expectations.
Prompting best practices for architects
Focus Each Diagram
"Focus only on Opportunity Management" produces a cleaner diagram than asking for the full system at once. Scope first, then expand.
Provide Schema Hints for SQL
Always include a one-line schema hint before asking for SQL optimization. "Table dbo.Orders(order_id PK, customer_id, total_amount)".
Specify the SQL Dialect
Always tell Copilot the SQL dialect: Oracle, SQL Server, or PostgreSQL. Optimization strategies differ significantly between them.
Iterate on Diagrams
If the first diagram isn't right, say "Add the data layer" or "Simplify the integration section." Stay in the same session to preserve context.
Use Coding Agent Early
Add Copilot as a reviewer as soon as a PR is opened — not just before merge. Early automated review catches issues before human reviewers spend time on them.
Validate Architecture Outputs
Always review generated diagrams against the actual requirements. Copilot may infer components not explicitly stated — validate every node and connection.
Why this matters
Step-by-step guide
Tools required