Ruthlessly tear apart code design — naming, abstractions, coupling, complexity, and everything else.
Generate Mermaid diagrams in the README to visualise architecture, flows, or relationships from the codebase.
Analyse test coverage gaps and report uncovered code before making changes.
Run tests, fix failures, and re-run until the suite passes.
Write and improve tests — reuse existing patterns, ensure consistency, and maintain quality.
Isolate a function or code block into a self-contained, runnable script for study and manual testing.
Fix grammar and improve writing while preserving the author's original voice and style.
Send plans, code, or approaches to OpenAI Codex CLI for independent review.