fcakyon/claude-codex-settings

What it solves

This project provides a collection of battle-tested configurations, plugins, and skills for AI coding tools like Claude Code, Codex CLI, Gemini CLI, and Cursor. It aims to fix common AI pitfalls such as over-engineering, bloated abstractions, and the use of machine-like buzzwords in generated text.

How it works

It functions as a plugin marketplace and library of extensions that can be installed into various AI CLI tools. These plugins implement specific behaviors through:

• Skills: Pre-defined capabilities (e.g., PDF processing, Python best practices) that the AI can use.
• Hooks: Scripts that run at specific triggers (e.g., blocking AI buzzwords before a file is saved or requiring a code review before a git commit).
• Output Styles: Instructions that modify how the AI formats its responses to be more concise and action-oriented.
• Agents: Specialized reviewer agents (like the Fable 5 advisor) for second opinions on big decisions.

Who it’s for

Developers using AI-powered coding assistants (Claude Code, Cursor, Codex CLI, Gemini CLI) who want to standardize their AI's behavior, enforce coding standards, and reduce the AI's tendency to overcomplicate code.

Highlights

• Simplify Plugin: Reviews changed code for redundancy and over-engineering across four angles (reuse, simplification, efficiency, altitude).
• Humanize Plugin: Uses a hook to block a curated list of AI buzzwords and cliches in markdown and docstrings.
• Intelligent Compact: Prevents Claude Code from losing critical information like file paths and root causes during session auto-summarization.
• Fable Advisor: Integrates a second-opinion reviewer agent for high-stakes decisions.
• Office Skills: Provides official Anthropic and OpenAI skills for processing PDF, Word, PowerPoint, and Excel files.
• ADHD Output Style: Reduces token usage by prioritizing answers and next steps over filler text.