Writing | peng

Writing

2026

Day 14：Mini Harness Recap：14 天架构复盘
Mini Coding Agent Harness 14 天系列第十四篇：复盘上下文、工具、安全与可观测性如何组成一个可检查的 Agent Harness。
Day 14: How to Build a Mini AI Coding Agent Harness in TypeScript
Day 14: connect context, tools, safety, and observability into one small, inspectable coding-agent harness.
Day 13：AGENTS.md：把项目规则注入 Agent 上下文
Mini Coding Agent Harness 14 天系列第十三篇：发现并注入 AGENTS.md 项目规则，同时保留来源与大小边界。
Day 13: AGENTS.md for AI Coding Agents: Load Project Rules into LLM Context
Day 13: discover project rules and add them to an agent context with bounded size and traceability.
Day 12：Approval 与 Sandbox：写入审批和工作区边界
Mini Coding Agent Harness 14 天系列第十二篇：用显式审批与工作区沙箱保护 Agent 的副作用。
Day 12: AI Coding Agent Security: Approval Gates and Workspace Sandbox in TypeScript
Day 12: protect coding-agent side effects with explicit approval gates and a workspace sandbox.
Day 11：Patch Editing：安全应用 Unified Diff
Mini Coding Agent Harness 14 天系列第十一篇：安全验证并应用多文件 Unified Diff，避免半完成写入。
Day 11: Patch Editing for AI Coding Agents: Apply Unified Diff Safely
Day 11: validate and apply multi-file unified diffs without partial writes.
Day 10：Git 与 Diff Tools：状态、补丁解析与变更摘要
Mini Coding Agent Harness 14 天系列第十篇：以只读方式让 Agent 查看 Git 状态、解析补丁并总结变更。
Day 10: Git Diff Tools for AI Coding Agents: Status, Parsed Patches, and Change Summaries
Day 10: give an agent safe, read-only visibility into repository status and parsed diffs.
Day 09：Context Explorer：调试 Prompt 与 Token 使用
Mini Coding Agent Harness 14 天系列第九篇：检查 Prompt、Token 预算与各类上下文的占用。
Day 09: Context Explorer: Debug AI Agent Prompts and Token Usage
Day 09: inspect an agent prompt, token budget, and the context parts that consume it.
Day 08：Context Builder：AI Coding Agent 上下文构建器
Mini Coding Agent Harness 14 天系列第八篇：把系统指令、用户请求、工具、规则与项目片段组装成可检查上下文。
Day 08: Context Builder for AI Coding Agents
Day 08: assemble system instructions, user requests, tools, rules, and project snippets into inspectable context.
Day 07：Transcript Logging：Agent 执行日志与可观测性
Mini Coding Agent Harness 14 天系列第七篇：用 JSONL 记录并回放 Agent 的模型请求、工具调用、结果与错误。
Day 07: Transcript Logging for AI Agent Runs
Day 07: record model requests, tool calls, results, errors, and timings as replayable JSONL transcripts.
Day 06：Agent Loop：Observe, Decide, Act / Coding Agent 执行循环
Mini Coding Agent Harness 14 天系列第六篇：把手动工具接入有边界的 observe-decide-act 执行循环。
Day 06: Agent Loop: Observe, Decide, Act in a Coding Agent
Day 06 of the Mini Coding Agent Harness series: turn manual tools into a bounded observe-decide-act loop.
Day 05：Local File Tools / 本地文件工具
Mini Coding Agent Harness 14 天系列第五篇：实现 list_files、read_file 和 search_text，让 harness 具备只读观察 workspace 的能力。
Day 05: Local File Tools
Day 05 of the Mini Coding Agent Harness series: implementing list_files, read_file, and search_text so the harness can observe a workspace safely.
Day 04：Tool Registry / 工具注册表
Mini Coding Agent Harness 14 天系列第四篇：定义工具 schema、runtime handler 和 dispatch，让 harness 具备可扩展的工具注册与手动调用能力。
Day 04: Tool Registry
Day 04 of the Mini Coding Agent Harness series: defining tool schemas, runtime handlers, dispatch, validation, and the first mock tool.
Day 03：LLM Chat Loop / 最小对话循环
Mini Coding Agent Harness 14 天系列第三篇：定义 chat message 和 model provider，把 CLI 从静态 scaffold 输出升级成最小模型对话循环。
Day 03: LLM Chat Loop
Day 03 of the Mini Coding Agent Harness series: defining chat messages and model providers, then upgrading the CLI from static scaffold output to a minimal chat loop.
Day 02：Project Scaffold / 项目骨架
Mini Coding Agent Harness 14 天系列第二篇：用 TypeScript 和 npm workspaces 搭出第一个可运行 CLI，让后续 agent loop、tools、context 和 transcript 有清楚的入口。
Day 02: Project Scaffold
Day 02 of the Mini Coding Agent Harness series: setting up a TypeScript and npm workspaces scaffold with the first runnable CLI surface.
什么是 Agent Harness：模型、Agent 和 Harness 的边界
Mini Coding Agent Harness 14 天系列第一篇：从 OpenAI 和 Anthropic 的真实接口形态出发，拆清模型、agent loop 和 harness 的边界。
What Is an Agent Harness: The Boundary Between Model, Agent, and Harness
Day 01 of the Mini Coding Agent Harness series: using real OpenAI and Anthropic API shapes to separate model APIs, agent loops, and the harness around them.
NuPhy 键盘快捷键速查笔记：系统切换、蓝牙、RGB 与 VIA 一次理清
入手 NuPhy 键盘后，我把官方使用指南里的关键快捷键整理成一份个人速查笔记，覆盖系统/连接切换、RGB 背光与侧灯、休眠、截图、恢复出厂以及 VIA 映射。原文见 nuphy.dev。
NuPhy Keyboard Shortcut Cheat Sheet: System, Bluetooth, RGB & VIA in One Pass (EN)
After picking up a NuPhy keyboard, I turned the official quick-start guide into a personal cheat sheet — covering system switching, device pairing, RGB backlight & sidelite, sleep, screenshots, factory reset, and VIA remapping. Original at nuphy.dev.

2025