AI-Agent-Expert: A Codex Expert-Level Skill System for Chinese Users

AI-Agent-Expert is a Codex expert-level Skill system designed specifically for Chinese users, aiming to fill the language barrier gap in the toolchain for Chinese AI developers. It corely provides three capabilities: AI Agent workflow design guidance, local long-term memory management, and daily ecosystem intelligence collection. Adopting a modular architecture, it supports Feishu delivery and Obsidian output, serving as a professional assistant for Chinese users in the AI Agent field.

With the popularity of AI programming assistants like OpenAI Codex and GitHub Copilot, Chinese users face language barrier pain points: most tools and resources prioritize English, creating obstacles from documentation to community discussions. The AI-Agent-Expert project was born to fill this gap, focusing on Chinese users' needs, integrating workflow design, memory management, and intelligence collection functions, and adapting to commonly used tools like Feishu and Obsidian.

1. Chinese-First Codex Usage Guidance

Provides a Chinese-written Skill instruction system (including usage guides and best practices), AI Agent workflow templates, and intelligence source scoring rules to help Chinese users quickly master advanced Codex features.

2. Local Long-Term Memory Management

Uses NDJSON format for structured memory storage, supports categorized retrieval, and all data is stored locally to protect privacy—addressing the core challenge of AI Agent memory management.

3. Daily Ecosystem Intelligence Workflow

Automatically collects AI ecosystem dynamics (papers, open-source projects, etc.), generates structured daily reports, and supports Feishu delivery and Obsidian output.

Configuration System

Three-layer architecture: shared configurations (sources.toml for intelligence sources, memory.toml for memory, delivery.example.toml for delivery examples); local configurations require copying delivery.example.toml to delivery.toml and setting up personal Feishu/Obsidian information to protect privacy.

Installation Mechanism

Create a repository entry in the Codex Skill directory via the script python scripts/install_skill.py; updates take effect in real time without reinstallation.

Intelligence Workflow Execution

The script run_daily_intel.py supports dry run (--dry-run) and formal run, reading configurations, collecting intelligence, generating reports, and delivering them.

Personal Developer Efficiency Improvement

Quickly master advanced Codex usage, reuse AI Agent workflow templates, and maintain context memory across sessions.

Team Knowledge Management

Unify workflow standards, share intelligence to Feishu groups for synchronized technical dynamics, and沉淀 team best practices into Skill instructions.

AI Ecosystem Intelligence Monitoring

Customize monitoring sources (GitHub/RSS, etc.), automatically generate daily reports, and archive to Obsidian to build a personal knowledge base.

Chinese-First

Established the Chinese-first principle from the design phase: Chinese Skill instructions, default Chinese intelligence sources, support for Chinese content output, and deep customization for Chinese users' habits rather than simple translation.

Privacy Protection

Distinguish between shareable content (Skill instructions, configuration templates) and local content (personal configurations, memory data, intelligence snapshots). Local content is excluded via .gitignore, so users can safely sync the repository.

Status: Core functions have been implemented, including the basic Skill instruction system, memory management framework, daily intelligence workflow, and Feishu/Obsidian support. Future Directions: Plan to add more AI Agent workflow templates, intelligence source adapters, intelligent retrieval enhancements, team collaboration features, and integration with other Chinese AI tools.

AI-Agent-Expert is a deeply customized tool for Chinese AI developers. It is not only a Codex Skill but also a complete auxiliary ecosystem covering usage guidance, memory management, and intelligence collection. The modular architecture (separation of shared and private content, separation of configuration and code, etc.) ensures ease of use and maintainability, providing a low-threshold entry for Chinese developers to master Codex and AI Agent technologies and stay sensitive to the AI ecosystem. Its open-source nature supports community co-construction, making it a localized technical partner for Chinese developers.