我专注于 Codex、AI Agent 工具、自动化工作流和证据驱动的产品交付。
I focus on Codex, AI agent tooling, automation workflows, and evidence-led product delivery.
- 设计和维护 Codex Skills、Agent 指令与渐进式上下文
- Design and maintain Codex Skills, agent instructions, and progressive context
- 构建本地自动化、工具连接和多 Agent 协作流程
- Build local automation, tool integrations, and multi-agent workflows
- 研究 AI 产品、开发者工具与真实工作流的结合方式
- Explore how AI products and developer tools can work in real workflows
- 用代码、文档、界面状态和验证结果交付可复核成果
- Ship reproducible outcomes with code, documentation, UI state, and verification evidence
我会把“看起来完成了”和“确实完成了”区分开:先确认现状和范围,再做最小必要修改,最后检查实际运行状态。
I distinguish “looks complete” from “is verified”: inspect the current state and scope first, make the smallest necessary change, then validate the real result.
- codex-instruction-auditor:基于 OpenAI Developers 官方文章《Rethinking skills and prompts for GPT-6 Astra》的个人实现,用于审计和整理 Codex 的
AGENTS.md、Skills、Prompt 和失效链接。 - codex-instruction-auditor: a personal implementation inspired by OpenAI Developers’ Rethinking skills and prompts for GPT-6 Astra, for auditing Codex
AGENTS.md, Skills, prompts, and broken links.
Codex · AI Agents · Prompt Engineering · Developer Tools · Automation · Evidence-led Delivery
欢迎通过 GitHub Issues 或 Discussions 交流 Codex Skills、AI Agent 工具链和个人开发环境建设。
Feel free to open an issue or discussion about Codex Skills, AI agent tooling, and personal developer environments.