| name | architecture-review |
| description | Review an existing Python RAG codebase against an explicit RAG architecture document and produce a production-readiness backlog with priorities, rationale, and execution guidance. |
Architecture Review Skill
Goal
Produce a production-grade architecture review document for a non-trivial Python RAG system.
The output must be a prioritised, implementation-ready task backlog that a senior engineering team can directly execute to harden the system for production.
The review must explicitly evaluate:
- The current implementation
- The stated architecture document (
RAG_ARCHITECTURE.md)
- Known, explicitly deferred architectural areas
When to Use This Skill
Use this skill when the user asks to:
- Review a RAG or LLM system for production readiness
- Audit architecture compliance
- Identify gaps, risks, or missing safeguards
- Produce a structured backlog for engineering execution
Instructions
- Assume the system is non-trivial and already deployed in at least one environment.
- Assume the audience is senior engineers.
- Evaluate the codebase against:
- Production engineering standards
- The architecture document
- Explicit architectural constraints (including deferred areas)
- Produce one table per category, in the exact order specified below.
- Each task must be concrete, opinionated, and actionable.
- Do not explain code line-by-line.
- Focus on what must be done, not how the system generally works.
Categories
(Create EXACTLY one table per category, in this order)
- Architecture & System Design
- RAG Pipeline (Ingestion, Retrieval, Reranking, Generation)
- Prompting & Grounding Controls
- Code Quality & Python Practices
- Observability & Operations
- Evaluation & Quality Measurement
- Security, Privacy & Data Risks
- Performance, Latency & Cost
- Testing Strategy
Table Format (MANDATORY)
Each table must contain EXACTLY the following columns, in this order: