Codex Basic Skills EP02: Must-Have Plugins and Skills for Beginners - Build Your Personal AI Assistant, Codex Basic Tutorial and Practice

你會學到

• Codex 提供多種外掛程式（如瀏覽器、文件處理）和技能（如AI生圖、技能創造器），是擴展其功能與效率的基礎。
• 透過「Computer Use」外掛，AI 能自主上網搜尋、下載資料，實現背景自動化作業，大幅提升工作彈性。
• 將重複且流程固定的任務封裝為「Skill」，是利用 AI Agent 提升效率的核心方法，可自動執行複雜指令串。
• 有效管理 AI Agent 的上下文視窗和 Token 消耗，需一次性給予完整指令，並利用 Agents.md 記錄專案進度以優化溝通成本。
• 啟用「自動審核」功能可讓 AI Agent 在無需人類介入下持續工作，加快任務完成速度。

教學金句

「你只要是要他做的事情是一個固定的格式跟流程。然後最後輸出的結果，你只要結果就好。那這樣方式就很適合做 Skill。」 「你不要一次叫他做一件，你要一次把你要他做的事情一次把他送出這樣子最省 Token 的數量。」 「如果你人沒有在電腦前面的話，他就不會往下做。所以一般來說我都會選用自動審核。」 「要不然如果說你沒有任何的紀律或節制的話，你很容易就會很快的把你的用量把它用完。」 「重要的事情在這個專案資料解裡面會記錄到 agent.md 這個檔案裡面來。新的對話會重讀這個檔案，來知道說我們上次做到哪邊。」

適用情境

這部影片非常適合納入 AI 工具應用或 AI Agent 進階操作的課程模組中，特別是針對已對 Codex 或其他 AI Agent 有初步了解，並希望提升工作效率、優化資源使用的中高階學員。學員應具備基礎的 AI 概念，並理解自動化在工作流程中的價值。本影片旨在引導學員掌握外掛與技能的實用配置、利用 AI 進行複雜任務的策略性指令給予、以及透過專案管理機制（Agents.md）有效控制成本與保持 AI 性能的關鍵方法。

延伸主題

• Codex 外掛程式進階客製化開發與 API 串接。
• 提示詞工程最佳實踐：如何設計彈性且強大的 Skill 提示詞。
• 多 Agent 協同工作模式與專案管理案例分析。
• Codex 成本效益分析與 Token 消耗深度優化策略。
• 將 Codex 與其他生產力工具（如 Obsidian、GitHub）整合的實作教學。

---en---

What You’ll Learn

• Codex offers various plugins (e.g., browser, document processing) and skills (e.g., AI image generation, skill creator) which are fundamental for expanding its functionality and efficiency.
• Through the “Computer Use” plugin, AI can autonomously search and download data from the internet, enabling background automated operations and significantly increasing work flexibility.
• Encapsulating repetitive and standardized tasks into “Skills” is a core method for enhancing efficiency with AI Agents, allowing for automatic execution of complex command sequences.
• Effective management of the AI Agent’s context window and token consumption requires providing comprehensive instructions in a single prompt and utilizing Agents.md to log project progress for optimized communication costs.
• Enabling the “Auto-approve” feature allows the AI Agent to continue working without human intervention, speeding up task completion.

Teaching Quote

“If what you want it to do follows a fixed format and process, and you only need the final output, then this approach is very suitable for creating a Skill.” “Don’t ask it to do one thing at a time; send all your requests at once to save the most tokens.” “If you’re not in front of the computer, it won’t proceed. So generally, I choose auto-approve.” “Otherwise, if you lack discipline or control, you’ll quickly use up your allocation.” “Important information for this project will be recorded in the agent.md file within the project directory. New conversations will reread this file to know where we left off last time.”

When to Use

This video is highly suitable for inclusion in AI tool application or advanced AI Agent operation course modules, especially for intermediate to advanced learners who have a preliminary understanding of Codex or other AI Agents and wish to improve work efficiency and optimize resource usage. Learners should possess basic AI concepts and understand the value of automation in workflows. This video aims to guide learners in mastering practical plugin and skill configurations, strategic command-giving for complex tasks using AI, and key methods for controlling costs and maintaining AI performance through project management mechanisms (Agents.md).

Related Topics

• Advanced custom plugin development and API integration for Codex.
• Best practices for prompt engineering: how to design flexible and powerful Skill prompts.
• Multi-Agent collaborative work models and project management case studies.
• Codex cost-benefit analysis and deep optimization strategies for token consumption.
• Practical tutorials on integrating Codex with other productivity tools (e.g., Obsidian, GitHub).