Agent 光会 grep 还不够:CodeGraph 让它先看懂代码结构
CodeGraph 的价值不是替代 grep,而是让 Agent 在真实代码库里先看定义、调用方、被调用方和影响面,减少盲搜、误判入口和漏改调用链的问题。
I build AI engineering tools, developer workflows, and knowledge systems for serious software work.
Public tools shaped around agents, database investigation, and skill distribution.
A native iOS client for linux.do: a SwiftUI app that renders the Discourse community natively while punching through Cloudflare and keeping you logged in.
The package manager for AI agent skills. Install, update, sync, and govern skills across coding agents.
Browse the real connections between posts, tags, and knowledge notes.
Long-running notes on tool building, engineering efficiency, and Obsidian workflows.
Notes on CodeGraph, skill-hub, agent frameworks, and AI-assisted development practice.
Backend systems, development environments, deployment entry points, and engineering toolchain practice.
How agents, Obsidian, and knowledge graphs can form a durable knowledge workflow.
Recent writing on agent coding, engineering workflows, and knowledge systems.
CodeGraph 的价值不是替代 grep,而是让 Agent 在真实代码库里先看定义、调用方、被调用方和影响面,减少盲搜、误判入口和漏改调用链的问题。
Agent 与 Obsidian 结合后,个人知识库不再只是静态笔记仓库,而可以成为可检索、可整理、可复用、可输出的长期知识工作台。
深入对比 OpenClaw 和 Hermes Agent 两个开源 AI Agent 框架,探讨它们在渠道覆盖、记忆系统、插件生态等方面的设计差异与适用场景。