codex-my-workflow

This repository is a Codex-focused port of the original claude-code-my-workflow project by Pedro Sant'Anna.

It keeps the core idea of the original repo: treat an AI coding assistant less like a chat partner and more like a disciplined contractor for academic and research work. The assistant should plan before non-trivial work, preserve state on disk, default to specialist review when it fits, use adversarial QA for substantial Beamer/Quarto parity work, verify outputs, and summarize what actually happened.

What changed here is the execution surface. The original repo was built around Claude Code. This version adapts that workflow to Codex by replacing Claude-specific mechanisms with Codex-native guidance, skills, agents, hooks, and durable artifacts.

Credit And Provenance

The upstream source for this repo is Pedro Sant'Anna's claude-code-my-workflow. This repository is a cloned and adapted Codex-first port, not a claim of authorship over the original workflow design.

The original project contributed the key workflow ideas:

• plan-first execution for non-trivial work
• specialized reviewers instead of one general reviewer
• adversarial critic/fixer loops for Quarto-vs-Beamer QA
• quality thresholds for deciding when work is ready
• durable workflow memory through plans, logs, and reusable guidance

This port keeps those ideas while translating them into the primitives Codex actually exposes.