AI Review

This repo contains carefully crafted context that you can load into an agentic LLM copilot like VS Code or Cursor. The agent then performs in specialized roles to give you exactly the help you want.

Finding problems

1. Start a new chat in the copilot for fresh context.

2. Begin the chat with something like:

   Read the file ai-review/critic/relational.md and
   critique the file myschema.sql

The copilot will read the files and then give you a very focused review of in this case the soundness of relational data modeling in your file. Copilots are also able to critique selected text in the editor if you say “critique the selected text.”

Fixing problems

After running a critic, you can switch modes and work on fixing the problems.

Use a prompt like

Fix the problems identified by the critic, using the
ai-review/fixer/conservative.md strategy