claude-codex-settings

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

"They make wrong assumptions on your behalf and just run along with them without checking. They don't manage their confusion, they don't seek clarifications, they don't surface inconsistencies, they don't present tradeoffs. They really like to overcomplicate code and APIs, they bloat abstractions, they don't clean up dead code after themselves." -- Andrej Karpathy

This repo's guidelines are structured to fix exactly these pitfalls.

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

Codex installs plugins from a local marketplace rather than a direct shell install command.

1. Clone this repo locally and open it in Codex.
2. This repo already includes .agents/plugins/marketplace.json.
3. If Codex was already open when you added or changed that file, restart Codex.
4. In Codex, open /plugins.
5. Choose Claude & Codex Settings and install the plugins you want.

This README only covers installing this marketplace. For generic Codex marketplace examples and maintainer docs, see CLAUDE.md.

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

Spawn a stronger Fable 5 reviewer to pressure-test a plan or conclusion before you commit, a drop-in for the built-in advisor when the Opus-main plus Fable-advisor pairing fails with a bare "unavailable" (#73365). It checks your load-bearing claims against the actual files and returns a skeptical review, not a rewrite.

Agents:

• fable-advisor - Fable 5 second-opinion reviewer, pass it the task, your approach, and the evidence

When Claude Code auto-summarizes a long session, the default summary routinely drops the highest-signal facts. This plugin tells the summarizer to keep them:

• File paths under investigation so the next turn doesn't re-discover where you were
• Confirmed root causes so you don't re-debug what's already solved
• Open questions, metrics, and IDs that prose summaries usually round away
• Findings from expensive subagent runs that took minutes to gather

Runs on every /compact (manual) and every auto compaction. Claude Code only; Codex, Cursor, and Gemini CLI don't yet expose a comparable summary hook.

Hooks:

• precompact_priorities.sh - Priority-preservation instructions for the compaction summarizer

Adds the two missing pieces Claude Code's telemetry needs to power a usage dashboard:

• Sticky session ID per project: resumed conversations stay one session, not dozens
• Categorized rejection reasons (profanity, wrong target, scope drift, retry, and more): chart why Claude pushes back

Per-device data is already in Claude Code's built-in OpenTelemetry stream. Pairs naturally with openobserve-skills for the dashboard side.

Hooks:

• session_start_chat_id.py - SessionStart hook that emits a sticky per-project chat_id
• user_prompt_reject_feedback.py - UserPromptSubmit hook that categorizes tool-rejection reasons

Skills CLI

npx skills add https://github.com/fcakyon/claude-codex-settings/tree/main/plugins/anthropic-office-skills --skill '*'

Official office document skills from anthropics/skills.

Skill	Description	Install
pdf	PDF processing (read, merge, split, create, OCR, forms)
pptx	PowerPoint presentation building and editing
xlsx	Excel spreadsheet processing with formulas	[](https://github.com/fcakyon/claude-codex-settings/releases/latest/do