codex-claude-code-config plugin
A plugin and kit for Claude Code and Codex: agent working principles, safety hooks, skills and CLAUDE.md templates in one repository
- 214 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The installer appends hooks to `~/.claude/settings.json` and `~/.codex/hooks.json`, and they fire on agent actions across all of the user's projects
- `sync_skills_to_codex.py --apply` copies skills into the Codex, Claude Code and `~/.agents` directories, overwriting changed local files after a backup
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/codex-claude-code-configDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The installer appends hooks to `~/.claude/settings.json` and `~/.codex/hooks.json`, and they fire on agent actions across all of the user's projects
- `sync_skills_to_codex.py --apply` copies skills into the Codex, Claude Code and `~/.agents` directories, overwriting changed local files after a backup
This is third-party code. Review the repository files before installing.
What it does
codex-claude-code-config is a configuration kit for Claude Code, Codex and other coding agents. The repository gathers about thirty working principles for agents (separate generator and evaluator agents, mandatory proof artifacts, defense against malicious packages and injections), Python hooks, ready rule files for `.claude/rules/`, `CLAUDE.md` and `REVIEW.md` templates, a few dozen skills and dynamic workflow commands such as `/deep-review-flow`. It installs one of three ways: as a Claude Code plugin, globally with the `install_hooks.py --global` script, or locally inside a project. Five safety hooks (dangerous commands, secret leaks, destructive Git operations, Git auto-backup, session drift check) form the minimum set for any project. After installing, the agent reads the repository's `AGENTS.md` and picks principles, hooks and skills for your project. Separate notes in `docs/` analyze the incidents each rule grew out of: token overspend on subagents, long waits and the one-hour cache, term drift in agent output.
Who it is for. Developers with long-running projects in Claude Code and Codex who want ready rules, safety hooks and CLAUDE.md templates.
Good fit when
- The agent reports the task as done while verification proves nothing: the "Proof Loop" principle requires saving evidence of the result
- Parallel chats share one machine and overwrite each other's state: the session coordination principle introduces lock files and handoff logs
- New dependencies keep arriving and you want protection from fresh malicious packages: two configuration lines from the supply chain principle block packages younger than seven days
- A session waits on CI or model training for more than an hour and the cache goes cold: the `long-wait-cache-ttl` rule caps waiting at fifty minutes
- A new project needs a starter `CLAUDE.md` for its type: templates sit in the `templates/` folder
Not a fit when
- You only need one small skill: the repository is large and installing it for a single file is excessive
- Your working environment is only Cursor or Windsurf: the plugin and hook installer are described for Claude Code and Codex
- You need a configuration with your team's real host names and secret files: the kit is a sample and has to be adapted
- Global hooks on a shared work machine are not agreed yet: the installer changes agent settings for all of the user's projects
Example request
Read AGENTS.md in the configuration repository and pick safety hooks and rules for my Next.js projectLimitations
- The `rules/` tree is a sample and nothing in it loads on its own: copy the needed files into the project's `.claude/rules/`.
- The hook installer appends entries to `~/.claude/settings.json` and `~/.codex/hooks.json`, so review the hook list before a global install.
- Some skills are narrow, for example for ML pipelines and diffusion models, and are useless in an ordinary project.
- The materials are in English; the README has only a short summary in Russian.
How to disable. Remove the plugin through Claude Code's plugin manager (`/plugin`). Delete the hook entries from `~/.claude/settings.json` and `~/.codex/hooks.json` and the copied skills from `~/.claude/skills`.
FAQ
How do I install the configuration as a Claude Code plugin?
Run `claude plugin marketplace add AnastasiyaW/codex-claude-code-config`, then `claude plugin install harness@aw-harness`. The plugin brings skills, agents, workflows, safe-default hooks and an index of rules injected at session start.
Which hooks are in the minimum set?
Five: `destructive-command`, `secret-leak`, `git-destructive`, `git-auto-backup` and `session-drift-validator`. `install_hooks.py` installs them: the `--global` flag for all projects, `--local` for the current one.
How do I install the kit into a single project?
Clone the repository into `.claude-config`, run `python .claude-config/scripts/install_hooks.py --local` and copy `skills` into `.claude/skills`. Global settings stay as they were.
Can I use the kit with Codex?
Yes. `python scripts/sync_skills_to_codex.py --apply --also-claude --also-agents` syncs public skills to Codex, `~/.agents` and Claude Code and makes backups of changed local files.
What if skills vanished after moving to a new machine?
Run `python scripts/recover_skill_trees.py --report` and read `docs/skill-tree-recovery.md`: no error appears, the skills just stop being offered.
Discussion
No questions or reviews yet.
Related
PluginsOne config for Claude Code and Codex in every repo: shared rules, skill packs, second-model review and a destructive-command guard
A Claude Code workspace folder: hooks with a local Ollama detect the project type, pick skills and remind you to update the docs
A Claude Code plugin marketplace with ready sets of agents, slash commands, hooks and skills, plus a search site
A template for Claude Code projects: parallel agent swarms, workflow skills, safety hooks and rules in one install.
A large kit for Claude Code: subagents, skills, commands, hooks, rules, templates and MCP configs in one plugin marketplace
A Claude Code reference plus a collection of over 70 skills for development, security, data and docs, with agent configs
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| codex-claude-code-config (this card) | no | 214 | |
| Trail of Bits Claude Code Config | no | 2,146 | Trail of Bits' set of recommendations without an install package or turnkey hooks |
| My Claude Code Setup | no | 2,717 | a starter kit with a CLAUDE.md template and a memory system, without principles or incident write-ups |
README in short
The README opens with notes on token overspend, long waits and result verification, then describes three install paths and a table of the minimum set for different project types. The main part pairs each common agent problem with a principle and a link, followed by the directory layout, maturity levels, supply chain protection, session handoff and a skills catalog. Short summaries in Chinese and Russian close it.