Help us improve
Share bugs, ideas, or general feedback.
A public skill collection centered on agent design philosophy inspired by Claude Code.
npx claudepluginhub dadwadw233/claude-code-philosophyA single-skill package for designing useful, controllable, extensible agentic systems.
Share bugs, ideas, or general feedback.
Design better agents.
claude-code-philosophy is a public skill and plugin package for developers who want to build agentic systems that are actually useful: controllable, legible, recoverable, and grounded in real product constraints.
It is inspired by the design philosophy behind Claude Code-class agent systems, but written as vendor-neutral guidance that can help builders working with Codex, Claude Code, or their own custom agent runtimes.
This project is unofficial. It is not affiliated with Anthropic.
A lot of agent projects are still too shallow:
This project exists to push in the other direction.
The core belief is simple: the difference between a toy agent and a useful one is usually the harness, not just the model or the prompt.
This repository currently ships one installable skill:
agent-design-coachIt helps developers:
The skill is built around a few strong principles:
Inside Codex, run:
$skill-installer install https://github.com/dadwadw233/claude-code-philosophy/tree/main/skills/agent-design-coach
Then restart Codex.
Inside Claude Code, run:
/plugin marketplace add dadwadw233/claude-code-philosophy
/plugin install claude-code-philosophy@claude-code-philosophy
Then invoke the installed skill with:
/claude-code-philosophy:agent-design-coach
~/.codex/plugins/claude-code-philosophy and register it in ~/.agents/plugins/marketplace.jsonskills/agent-design-coach to ~/.claude/skills/agent-design-coachDetailed examples are below.
This is the simplest path if you only want the skill itself.
$skill-installer install https://github.com/dadwadw233/claude-code-philosophy/tree/main/skills/agent-design-coach
After installation, restart Codex.
Clone the repo:
git clone git@github.com:dadwadw233/claude-code-philosophy.git
Copy it into your local plugin area:
mkdir -p ~/.codex/plugins
cp -R /absolute/path/to/claude-code-philosophy ~/.codex/plugins/claude-code-philosophy
Create or update ~/.agents/plugins/marketplace.json:
{
"name": "personal-plugins",
"interface": {
"displayName": "Personal Plugins"
},
"plugins": [
{
"name": "claude-code-philosophy",
"source": {
"source": "local",
"path": "./.codex/plugins/claude-code-philosophy"
},
"policy": {
"installation": "AVAILABLE",
"authentication": "ON_INSTALL"
},
"category": "Productivity"
}
]
}
Restart Codex. The plugin should appear in the plugin directory.
This repository includes a Claude Code marketplace manifest at .claude-plugin/marketplace.json.
Register the marketplace:
/plugin marketplace add dadwadw233/claude-code-philosophy
Install the plugin:
/plugin install claude-code-philosophy@claude-code-philosophy
After installation, the explicit slash command is:
/claude-code-philosophy:agent-design-coach
Copy the skill directory:
mkdir -p ~/.claude/skills
cp -R skills/agent-design-coach ~/.claude/skills/agent-design-coach
For a manually installed standalone skill, invoke:
/agent-design-coach
Use $agent-design-coach to turn this coding-agent idea into a real system architecture.Use $agent-design-coach to review this repo's agent design for harness, memory, permissions, and UX gaps./claude-code-philosophy:agent-design-coach/claude-code-philosophy:agent-design-coach Review this coding agent and tell me what is weak in its harness design.