Installation

Install CodexKit locally, then publish the docs when the pack matches your workflow.

CodexKit follows the official Codex Skills layout: each skill ships with SKILL.md, optional agents/openai.yaml, and standard .agents/skills discovery paths. The pack now covers engineering, high-reasoning work, low-reasoning office automation, department templates, and starter workspaces across project, finance, legal, operations, HR, strategy, analytics, marketing, and CX workflows.

⭐ View on GitHub Download latest release

Reference: OpenAI Codex Skills documentationRequires Node 20.9+ and npm 10+

Clone from GitHub

git clone https://github.com/hoavdc/CodexKit.git
cd CodexKit

This is the recommended way to get CodexKit. After cloning, follow the quick install steps below.

Fastest path for non-coders

1. Open the GitHub Releases page and download codexkit-starter-pack-v0.7.0.zip.
2. Unzip the package anywhere on your machine.
3. On Windows, double-click START-HERE-WINDOWS.cmd. On macOS or Linux, run bash ./START-HERE.sh.
4. Restart Codex and type /skills to confirm all 81 skills appear.
5. Open skill-finder.md to browse skills by situation.
6. Optional: create a starter workspace with CREATE-WORKSPACE-WINDOWS.cmd or bash ./CREATE-WORKSPACE.sh.

Prerequisite: Codex CLI must be installed. If not, run npm install -g @openai/codex first. The START-HERE scripts will detect and notify you if Codex is missing.

Windows PowerShell

.\scripts\install-skills.ps1
.\scripts\install-skills.ps1 -Destination .\.agents\skills
.\scripts\quick-start.ps1 -List
.\scripts\quick-start.ps1 -Starter project-management-office -Destination .\acme-pmo
node ./scripts/validate-pack.mjs

The default command installs into $HOME/.agents/skills. The second command installs repo-local skills into .agents/skills.

macOS / Linux

bash ./scripts/install-skills.sh
CODEXKIT_DESTINATION=./.agents/skills bash ./scripts/install-skills.sh
bash ./scripts/quick-start.sh --list
bash ./scripts/quick-start.sh --starter finance-performance-desk --destination ./acme-finance
node ./scripts/validate-pack.mjs

Use --force or CODEXKIT_FORCE=1 only when you want to overwrite an existing installed skill folder.

Updating CodexKit

CodexKit ships with update scripts that auto-detect your install method and pull the latest skills without re-downloading everything.

Git users

bash ./scripts/update-codexkit.sh

Windows

.\scripts\update-codexkit.ps1

Or double-click UPDATE-WINDOWS.cmd

Git clone: Runs git pull + re-installs all skills with --force.
Zip download: Fetches the latest release from GitHub, extracts new skills, and overwrites installed ones.

Release packaging

GitHub tags matching v* trigger the release workflow.
Each release publishes both a full source archive and a starter pack archive meant for end users.
The starter pack now includes skills, templates, workspaces, playbooks, automations, MCP guides, scripts, and quick-start launcher files.

Discovery paths

Codex scans .agents/skills from your current working directory up to the repository root.
Codex also scans $HOME/.agents/skills for user-wide skills that should follow you across repos.
If a newly installed or updated skill does not appear, restart Codex.

Skill structure

my-skill/
|-- SKILL.md
|-- agents/
|   `-- openai.yaml
|-- references/
|-- scripts/
`-- assets/

Starter workspace adoption

Choose the closest starter in workspaces/ instead of beginning from an empty folder.
Keep the structure, but rename files and sections to match your team’s actual cadence and terminology.
Pair the workspace with one high-reasoning skill and one routine automation skill before adding more assets.

For Codex users

Use $HOME/.agents/skills for personal defaults and .agents/skills when a team wants repository-scoped skill discovery that travels with the codebase.

For ChatGPT users

Use playbooks, automations, templates, and starter workspaces as copy-ready assets. The project still adds value even when local skill installation is unavailable.