ChatGPT Prompt for AI Code Review & Refactoring
Have Claude Code critique the architecture of a CLI wrapper around an LLM against stated goals and propose scoped improvements.
More prompts for AI Code Review & Refactoring.
Run an opinionated tech-debt audit on a SvelteKit app, prioritized by cognitive complexity and readability, producing a ranked backlog.
Have Continue.dev critique the architecture of a Python data pipeline against stated goals and propose scoped improvements.
Run an opinionated tech-debt audit on a monorepo (Nx), prioritized by cognitive complexity and readability, producing a ranked backlog.
Have Windsurf critique the architecture of a SvelteKit app against stated goals and propose scoped improvements.
Run an opinionated tech-debt audit on a Terraform IaC module, prioritized by database migration safety (online, reversible), producing a ranked backlog.
A safe, step-by-step refactor recipe for applying extract a reusable React hook in a Ruby Astro static site without breaking the test suite.
Acting as a world-class staff engineer consultant, you provide strategic guidance to CTOs in the architecture review space. Critique the architecture of a CLI wrapper around an LLM using **Claude Code** as the driving agent. Output a principled, evidence-based critique with scoped recommendations -- not a greenfield redesign. **Project:** CLI wrapper around an LLM **Team:** solo founder **Driver tool:** Claude Code **Stated quality goal:** enable 3x team growth without onboarding pain **Convention baseline:** no implicit env var reads -- centralize in config.ts ## Stated goals you are evaluating against 1. enable 3x team growth without onboarding pain 2. Respects no implicit env var reads -- centralize in config.ts 3. Safe to evolve (tests, rollbacks, feature flags) 4. Reasonable cost and operability ## Artifacts to ingest - Repo root files (README, CLAUDE.md, package/requirements manifests) - Top-level directory layout (3 levels) - Module-dependency graph (you may approximate from imports) - Deployment topology (Dockerfile, IaC, CI) - Request path for the primary user flow (describe it end-to-end) - Data model (key entities + relationships) - Any existing ADRs If these are missing, ask once -- then proceed with the best read you have. ## Critique output ### 1. Architecture snapshot (one page) - Pattern name (monolith / modular monolith / microservices / serverless / hybrid) - Core subsystems and their responsibilities (one line each) - Primary data store(s) and their access pattern - Deployment unit(s) and their SLO (if stated) - External dependencies (hard and soft) ### 2. Strengths (3-5 items) For each: what's working, why it fits enable 3x team growth without onboarding pain, and a single piece of evidence. ### 3. Weaknesses (5-8 items), ranked For each: - **Finding** (one sentence) - **Why it matters for enable 3x team growth without onboarding pain** (one sentence) - **Evidence** (specific file, pattern, or metric) - **Severity** (critical / high / medium / low) - **Cost of inaction** (quantified if possible -- latency, incident MTTR, onboarding time) ### 4. Coupling and boundary review - List the 5 most problematic cross-module dependencies - For each, propose the smallest intervention that breaks the coupling (introduce an interface, move code, add a queue) ### 5. Data layer review - Read vs write patterns and how the schema supports or fights them - 3 queries that will not scale past `<threshold>` and why - Migration safety posture (online? reversible? shadowed?) ### 6. Operability review - Observability gaps (per-request trace? structured logs? RED metrics?) - Failure modes you can name and the blast radius of each - Runbook coverage ### 7. Scoped recommendations A ranked list of 6-10 changes, each with: - Title, effort (S/M/L), risk (low/med/high) - The 1-paragraph sequence to execute it safely - Success metric with a target number - Prerequisites ### 8. Non-recommendations (explicit) 3-5 things you considered and rejected, with reasons. (Big rewrites, trendy tech swaps, speculative generality.) ### 9. 90-day plan Map the top recommendations onto a 90-day calendar with weekly milestones. Each milestone is a testable outcome. ## Style rules - No "industry best practice" without naming which industry and which practice - No rewrites -- only refactors and targeted introductions - Every claim is evidence-linked - Respect no implicit env var reads -- centralize in config.ts as a design constraint, not a thing to renegotiate Structure as a professional report with: Executive Summary, Key Findings, Detailed Analysis, Recommendations, and Next Steps.