Research prototypes, tools and engineering projects. Dates are GitHub repository creation months, not the actual project periods.
Research prototype
A prototype that supervises the tool calls of LLM agents. It places the security boundary after the agent emits a tool call and before the tool runs, auditing, scoring, blocking and logging file access, command execution, API calls and outbound actions, and it detects cross-step chains such as “read sensitive data → send externally”.
Research prototype
An experience-driven system for evolving multi-agent organisations, built on openJiuwen Core. It runs tasks with the current Strategy and only proposes candidate strategies offline when evidence such as repeated failures crosses a threshold; a candidate changes the default organisation for later tasks only after independent validation.
Tool
A Markdown sticky note for the desktop that stays on screen, collapses into a small capsule and opens with a global shortcut. Installers are published on GitHub Releases; according to the README, the current development build has been verified on Apple Silicon macOS.
An agent skill for reading papers in depth: explaining methods and equations, checking evidence, filling in related work and keeping continuously updated notes. It offers nine modes such as quick, deep, critique and compare. MIT licensed.
Engineering project
A configurable CRM platform for multiple companies, aiming to turn business processes that small companies keep in Excel into an online system with staff accounts, permissions and statistics. It is a pnpm monorepo with separately deployed web, API and worker processes.
Research prototype
A repository for detecting AI-generated content and attributing it to source models, with modules for detection, provenance and data ingestion.
Research prototype
A prototype memory engine for agents in enterprise collaboration, intended to turn Feishu chats, notes and meeting discussions into structured memory with provenance. It is currently a local MVP: data models, SQLite storage and sample chat ingestion are done; memory extraction and retrieval are not implemented yet.
Competition project
An entry for the China Software Cup: an intelligent rich-text editor for writing, study and office work, with document management, AI-assisted writing, image OCR and question answering over a local knowledge base. Vue 3 + Tiptap front end, Flask + MongoDB back end.