← Back to list
codex Checked
**Generate images** with Codex's built-in OpenAI image-generation tool — route every "生成图片 / 配图 /…
8.3Overall rating
Installation
🤖 Install via AI
Copy the prompt below and send it to your AI assistant (e.g. Claude Code) — it will follow the instructions and install automatically.
Install the "codex" skill by following the instructions at https://skill123.me/install/codex.
⌨️ Command line install
Run in your terminal — downloads and installs to ~/.claude/skills/.
curl -fsSL https://skill123.me/install/codex.sh | bash
📦 Download ZIP
Download the zip and extract it into your skills directory (e.g. ~/.claude/skills/), then restart your session.
⬇ Download v1.0.0 · 10 KBAbout this skill
Overview
**Generate images** with Codex's built-in OpenAI image-generation tool — route every "生成图片 / 配图 / 插图 / image gen / 画一张 / 出一套图" request here; this is the primary use. ALSO runs Codex CLI as a background sub-agent for code analysis, refactoring, review, or an agent team working in parallel. Always runs in background; uses Codex's default model (no -m flag) unless the user explicitly overrides.
Source
- Repo: https://github.com/yan-labs/yan-skills
- Path: codex
Score breakdown
Trigger
7.0
Clear What and bilingual trigger keywords for image gen and sub-agent use, but no exclusion boundary and coverage beyond that is thin.
Structure
7.3
Compact 236-line SKILL.md with no offload needed, but README.md ships in the package.
Workflow
8.8
Strong task design: background launch protocol, team composition, collision rules, and a three-check image acceptance loop including blind review.
Content
9.0
Concrete runnable commands and imperative style; some long personal war stories dilute density.
Engineering
10.0
Valid frontmatter, no scripts to break, no dangling path references.
Security
8.5
No scripts or secrets; but it instructs agents that sandbox flags need no permission prompt, a gate-removal pattern risky for public distribution though disclosure is mandated.
Well-crafted tool-wrapper skill for Codex background workers and image generation with excellent verification practices; flaws are a bundled README, no When-Not boundary, and machine-specific advice to skip permission prompts for danger-full-access sandbox.