claude codex settings

My daily battle-tested Claude Code/Desktop and OpenAI Codex setup with skills, commands, hooks, subagents and MCP servers.

Installation • Plugins • Configuration • References

Installation

Plugins add skills, commands, and automations to your AI coding tool. Install only what you need from the plugin list below.

Prerequisites: See INSTALL.md for setup requirements.

# Add marketplace (one time)
/plugin marketplace add fcakyon/claude-codex-settings

# Install any plugin by name
/plugin install < plugin-name > @claude-settings

After installing, run /plugin-name:setup for configuration (e.g., /gcloud-tools:setup).

codex plugin install < plugin-name > @claude-settings

gemini extensions install --path ./plugins/<plugin-name>

cursor plugin install < plugin-name > @claude-settings

Create symlinks for cross-tool compatibility:

ln -sfn CLAUDE.md AGENTS.md
ln -sfn CLAUDE.md GEMINI.md

Plugins

Official office document skills from anthropics/skills. Synced locally via bash .github/scripts/sync-anthropic-office-skills.sh.

Skill	Description	Install
pdf	PDF processing (read, merge, split, create, OCR, forms)
pptx	PowerPoint presentation building and editing
xlsx	Excel spreadsheet processing with formulas
docx	Word document creation and editing

Official office document skills from openai/skills. Synced locally via bash .github/scripts/sync-openai-office-skills.sh.

Skill	Description	Install
pdf	PDF generation and extraction with visual review
slides	Slide deck creation with PptxGenJS
spreadsheet	Spreadsheet processing with formulas and formatting
doc	Word document creation and editing

React and frontend best practices synced from vercel-labs/agent-skills.

Skills (ZIP for claude.ai, Claude Code, Cursor, Codex, VS Code):

Skill	Description	ZIP
composition-patterns	React composition patterns: compound components, render props, context
react-best-practices	64 React/Next.js performance rules from Vercel Engineering
react-native-skills	35+ React Native/Expo rules for performance and animations
react-view-transitions	View Transition API for page/route animations
web-design-guidelines	UI review against 16 web interface guideline categories

Browser automation via CLI instead of MCP. 93% less context usage than Playwright MCP by using snapshot + element refs instead of full DOM tree dumps. Synced from vercel-labs/agent-browser.

Skills (ZIP for claude.ai, Claude Code, Cursor, Codex, VS Code):

Skill	Description	ZIP
agent-browser	Browser automation: navigation, forms, clicking, screenshots, auth, sessions
electron	Automate Electron desktop apps (VS Code, Slack, Discord, Figma, Notion)

CLI Tool: agent-browser - install via npm i -g agent-browser && agent-browser install

Frontend design skills from anthropics/claude-plugins-official and openai/skills. Synced locally via bash .github/scripts/sync-frontend-skills.sh.

Skill	Description	Install
openai-frontend-design	Composition-first design: restrained layout, image-led hierarchy, tasteful motion
anthropic-frontend-design	Bold aesthetic direction, distinctive typography, anti-generic AI aesthetics

Official MongoDB agent skills for schema design, query tuning, search, and connections. Synced from mongodb/agent-skills.

Skills (ZIP for claude.ai, Claude Code, Cursor, Codex, VS Code):

Skill	Description	ZIP
atlas-stream-processing	Atlas stream processing patterns
mongodb-connection	Connection management
mongodb-mcp-setup	MCP server setup
mongodb-natural-language-querying	Natural language to queries
mongodb-query-optimizer	Query performance tuning
mongodb-schema-design	Schema design patterns
mongodb-search-and-ai	Atlas Search and AI integration

Official Supabase Postgres best practices for query and schema design. Synced from supabase/agent-skills.

Skills (ZIP for claude.ai, Claude Code, Cursor, Codex, VS Code):

Skill	Description	ZIP
supabase-postgres-best-practices	Postgres performance and schema design across 8 categories

Best-of bundle from anthropics/claude-plugins-official. Cherry-picks skills, agents, and commands from multiple upstream plugins.

Skills:

• claude-md-improver - Audit and improve CLAUDE.md files across a codebase
• skill-creator - Create, improve, and benchmark Agent Skills with eval testing

Agents:

• code-architect - Architecture design from codebase patterns
• code-explorer - Deep codebase exploration and analysis
• code-reviewer - Code quality and review

Commands:

• /feature-dev - Guided feature development workflow
• /revise-claude-md - Capture session learnings into CLAUDE.md

Hooks:

• Stop (prompt) - Suggest CLAUDE.md updates after significant code changes
• SessionEnd (prompt) - Remind to capture learnings when session ends

Actively maintained plugin development toolkit from anthropics/claude-plugins-official. 7 skills, 3 agents, and guided plugin creation.

Skills: hook-development, mcp-integration, plugin-structure, plugin-settings, command-development, agent-development, skill-development

Agents: agent-creator, plugin-validator, skill-reviewer

Commands: /create-plugin - Guided plugin workflow

Academic research toolkit from fcakyon/phd-skills. Experiment design, literature review, paper writing, citation verification, and reviewer defense for PhD workflows.

Skills:

• dataset-curation - Dataset preparation and annotation guidelines
• experiment-design - Experiment setup and ablation planning
• latex-setup - LaTeX project configuration
• literature-research - Literature search and review
• paper-verification - Citation and claim verification
• paper-writing - Academic paper drafting
• research-publishing - Submission and publishing workflow
• reviewer-defense - Reviewer response preparation

Agents:

• experiment-analyzer - Analyze experiment results and suggest next steps
• paper-auditor - Audit paper for consistency and completeness

Commands:

• /factcheck - Verify citations and claims
• /fortify - Strengthen paper against reviewer critiques
• /gaps - Find gaps in literature coverage
• /setup - Configure phd-skills
• /xray - Deep analysis of a paper

Git and GitHub automation. Run /github-dev:setup after install.

Agents:

• commit-creator - Intelligent commit workflow
• pr-creator - Pull request creation
• pr-reviewer - Code review agent
• pr-comment-resolver - PR comment resolution

Skills:

• commit-workflow - Commit process and message format
• pr-workflow - PR creation workflow
• pr-comment-workflow - PR comment style and resolution
• setup - GitHub CLI troubleshooting

Commands:

• /commit-staged - Commit staged changes
• /create-pr - Create pull request
• /review-pr - Review pull request
• /resolve-pr-comments - Address unresolved PR comments
• /update-pr-summary - Update PR description
• /clean-gone-branches - Clean deleted branches
• /github-dev:setup - Configure GitHub CLI

Hooks:

• git_commit_confirm.py - Confirmation before git commit
• gh_pr_create_confirm.py - Confirmation before gh pr create

Auto-formatting hooks for Python, JavaScript, Markdown, and Bash.

Hooks:

• format_python_docstrings.py - Google-style docstring formatter
• python_code_quality.py - Python code quality with ruff
• prettier_formatting.py - JavaScript/TypeScript/CSS/JSON
• markdown_formatting.py - Markdown formatting
• bash_formatting.py - Bash script formatting

40+ Azure services with Azure CLI authentication. Run /azure-tools:setup after install.

Skills:

• azure-usage - Best practices for Azure
• setup - Troubleshooting guide

Commands:

• /azure-tools:setup - Configure Azure MCP

MCP: .mcp.json | microsoft/mcp/Azure.Mcp.Server

Commands for syncing CLAUDE.md and permissions allowlist from repository, plus context refresh for long conversations.

Commands:

• /load-claude-md - Refresh context with CLAUDE.md instructions
• /sync-claude-md - Sync CLAUDE.md from GitHub
• /sync-allowlist - Sync permissions allowlist

Hooks:

• sync_marketplace_to_plugins.py - Syncs marketplace.json to plugin.json

Logs, metrics, and traces. Run /gcloud-tools:setup after install.

Skills:

• gcloud-usage - Best practices for GCloud Logs/Metrics/Traces
• setup - Troubleshooting guide

Commands:

• /gcloud-tools:setup - Configure GCloud MCP

MCP: .mcp.json | google-cloud/observability-mcp

Code quality agent and utility hooks.

Agent:

• code-simplifier - Ensures code follows conventions

Hooks:

• enforce_rg_over_grep.py - Suggest ripgrep

Search papers across arXiv, PubMed, IEEE, Scopus, ACM. Run /paper-search-tools:setup after install. Requires Docker.

Skills:

• paper-search-usage - Best practices for paper search
• setup - Troubleshooting guide

Commands:

• /paper-search-tools:setup - Configure Paper Search MCP

MCP: .mcp.json | mcp/paper-search

Web search and content extraction. Run /tavily-tools:setup after install.

Skills:

• tavily-usage - Best practices for Tavily Search
• setup - Troubleshooting guide

Commands:

• /tavily-tools:setup - Configure Tavily MCP

Hooks:

• webfetch_to_tavily_extract.py - Redirect WebFetch to Tavily extract
• websearch_to_tavily_search.py - Redirect WebSearch to Tavily search
• tavily_extract_to_advanced.py - Upgrade Tavily extract depth

MCP: .mcp.json | tavily-ai/tavily-mcp

Configuration

Configuration in .claude/settings.json:

• Model: OpusPlan mode (plan: Opus 4.6, execute: Opus 4.6, fast: Sonnet 4.6) - source
• Environment: bash working directory, telemetry disabled, MCP output limits
• Permissions: bash commands, git operations, MCP tools
• Plugins: All plugins enabled

Configuration in .claude/settings-zai.json using Z.ai GLM models via Anthropic-compatible API:

• Main model: GLM-5-Turbo (dialogue, planning, coding, complex reasoning)
• Fast model: GLM-4.7-Flash (file search, syntax checking)
• Cost savings: 85% cheaper than Claude 4.6 - source
• API key: Get from z.ai/model-api

Run Claude Code with Kimi K2.5 via Anthropic-compatible API - source:

• Model: kimi-k2.5 - High-speed thinking, 256K context
• API key: Get from platform.moonshot.ai

export ANTHROPIC_BASE_URL=https://api.moonshot.ai/anthropic
export ANTHROPIC_AUTH_TOKEN="your-moonshot-api-key"
export ANTHROPIC_MODEL=kimi-k2.5
export ANTHROPIC_DEFAULT_OPUS_MODEL=kimi-k2.5
export ANTHROPIC_DEFAULT_SONNET_MODEL=kimi-k2.5
export ANTHROPIC_DEFAULT_HAIKU_MODEL=kimi-k2.5
export CLAUDE_CODE_SUBAGENT_MODEL=kimi-k2.5
export ENABLE_TOOL_SEARCH=false

Configuration in ~/.codex/config.toml:

• Model: gpt-5.4 with model_reasoning_effort set to "high"
• Provider: Azure via responses API surface
• Auth: Project-specific base URL with env_key authentication

Settings in .vscode/settings.json:

• GitHub Copilot: Custom instructions for automated commit messages and PR descriptions
• Python: Ruff formatting with auto-save and format-on-save enabled
• Terminal: Cross-platform compatibility configurations

TODO

Visual demos:

• Add before/after comparison slider images or GIFs for each plugin showing the value visually

Zero-MCP goal:

• Replace MCP-based plugins with CLI alternatives where possible (mongodb, tavily, gcloud, azure, linear, supabase, paper-search)

New plugins/skills:

• Payments: Stripe best practices, Polar integration
• Deployment: Dokploy deployment skill
• Frontend design: Anthropic + OpenAI frontend design skills (bundled as frontend-design-skills)
• Frontend: TanStack (Router, Query, Table, Form)
• Real-time: LiveKit voice/video agent skill
• Documents: Google Docs, PPTX, DOCX, Excel from OpenAI (in addition to Anthropic ones)
• Auth: Clerk, Firebase, Supabase Auth patterns
• Fullstack: FastAPI, NodeJS backends, Tailwind CSS v4, shadcn/ui, Sentry monitoring, Web Vitals

Static website:

• Publish plugin catalog site with search, category filtering, and per-tool install snippets

Other:

• Rename plugins to mongodb-skills, github-skills etc. for better UX
• Add worktree support to github-dev create-pr and commit-staged commands

References

• Claude Code - Official CLI for Claude
• Claude Code Plugins - Plugin format reference
• Anthropic Skills - Official skill examples
• OpenAI Codex - Official CLI for Codex
• Codex Plugins - Plugin format reference
• Gemini CLI - Official CLI for Gemini
• Gemini Extensions - Extension format reference
• Cursor Plugins - Plugin format reference
• AGENTS.md - Cross-tool agent specification
• Agent Skills - Open format for giving agents new capabilities

Thank you for the support!