I love coding and designing elegant systems.
- 🤖 Exploring: Long-Horizon Harness & Long-term Memory & Agent Infra(Sandbox、Computer Use)
- 🎓 Education: IoT @ HHU | Applying for an MS in CS @ HKU / CUHK / HKUST
- 🧑💻 Internship: prev @GIAN - agent development
- 🌐 Links: visit my blog here or reach me on
正在找 agent 相关实习,欢迎联系 18896680730@163.com
- YuBlog
- 黑白极简 Astro 开源博客主题
- RepoMirror
- 将 GitHub 仓库或仓库子目录镜像到本地文件夹树的 macOS 桌面应用
- Trajex
- Past Claude Code, Codex, and Pi sessions — queryable by your agent, browsable by you
- Pai
- miniPi
- skills - 我平时使用的 Skill 集合,涵盖开发、设计和研究
- understand-project - 从整体架构、运行形态和真实数据链路读懂项目,并指导源码阅读与二次开发
- pr-lifecycle - 把 PR 当作持续收敛的契约:覆盖 SPEC、设计、实现、review、测试消融与最终审查的 Agent Skill
- video2article - 把视频/播客重写成「阅读版本」文章的 Agent Skill:YouTube 字幕、小宇宙官方逐字稿、本地 ASR
- pi-deep-dive
- In-depth analysis and source-code exploration of pi-coding-agent
- pi-extensions - 我平时使用和推荐的 pi 扩展集合
- pi-repomap - A Tree-sitter-powered codebase structure awareness extension for Pi
- pi-tool-offloading
- Pi的上下文卸载扩展,将超大工具结果移至会话侧边文件,按需检索
- pi-zen-mode
- pi 的极简模式扩展
- pi-dynamic-workflows - A Claude Code-like Dynamic Workflows extension for Pi
pi
· 6 PRs
· Extensible AI coding agent and development toolkit
memU
· 6 PRs
· Memory infrastructure for AI agents
OpenViking
· 9 PRs
· Context database and infrastructure for AI agents
deer-flow
· 12 PRs
· Open-source super agent harness for research and coding
deer-workflow
· 1 PR
· AI-powered workflow automation for software development
| PR | Commit |
|---|---|
| #7 | feat: support Pi Coding Agent 0.84.1 |
maka
· 11 PRs
· Agent that keeps a complete record of everything it did
powercontext
· 9 PRs
· Context for work that humans and agents hand off and continue
letta-code
· 1 PR
· Memory-first coding agent built on Letta
| PR | Commit |
|---|---|
| #4205 | fix(agent): keep memoryScope on provider-not-supported subagent retry |
llm-space
· 1 PR
· Infrastructure and runtime components for LLM applications
| PR | Commit |
|---|---|
| #148 | fix(core): confine LocalFileSystem paths |
open-connector
· 3 PRs
· Open-source connector infrastructure for AI applications
| PR | Commit |
|---|---|
| #435 | fix(oauth): prune expired authorization states |
| #436 | fix(runtime): handle JSON media types case-insensitively |
| #439 | fix(providers): allow private network credential validation |
OpenKB
· 2 PRs
· Open-source knowledge base for AI agents and RAG
| PR | Commit |
|---|---|
| #59 | auto-propagate LLM_API_KEY to provider-specific env vars via config-driven detection |
| #79 | Eliminate duplicate wiki-tool implementations in skill/tools.py |
deepagents-in-action
· 4 PRs
· datawhale deepagents book
| PR | Commit |
|---|---|
| #127 | docs: 新增第 5 章子 Agent 委派与上下文隔离 Notebook |
| #129 | docs: add ch06 async subagent lifecycle notebook |
| #132 | docs: 澄清第 6 章 Agent Server、ASGI 传输与服务启动流程 |
| #125 | docs: 修正第 1、5 章技术描述 |
Unmerged PRs
| Project | PR | Title |
|---|---|---|
| memU | #690 | feat(retrieve): fuse BM25 with cosine via RRF |
| OpenViking | #4658 | feat(session): extract session skills even when no training cases are produced |
| #4660 | feat(bot): deliver send:// generated images on Telegram, Discord, and Slack | |
| #4730 | docs(retrieval): clarify score threshold and hotness semantics | |
| #4736 | feat(bot): process independent sessions concurrently | |
| #5338 | perf(reindex): parallelize resource directory vector admission | |
| pi | #8744 | feat(tui): add opt-in overlay selection exclusion |





