mocX
中文

Beijing University of Posts and Telecommunications · Beijing

mocX

GitHub @Zhong0118

Interested in the security and collaboration of LLM agents，Testing ideas with runnable prototypes and tools

I publish code on GitHub as Zhong0118 and keep technical notes on the MocLab blog. Most recent public repositories deal with LLM agents: supervising tool calls, evolving multi-agent teams, and desktop and agent tools I use day to day.

View projectsOnline CVGitHubMocLab

Research directions and related projects

How can tool calls generated by an LLM agent be checked for privilege abuse and multi-step risk chains before they actually run?

Related projectsAgentGuard-ChainAll directions

Research directions

Three directions covered by the public projects, each linked to repositories you can inspect.

More about me

  • 01

    Agent security

    How can tool calls generated by an LLM agent be checked for privilege abuse and multi-step risk chains before they actually run?

    A security gateway between tool-call generation and execution audits, scores, blocks and logs file, command, API and outbound actions.

    Related projects

  • 02

    Multi-agent collaboration

    Can the roles and collaboration strategy of a multi-agent team change only after cross-task evidence and validation support it?

    EvoTeam constrains evolution with versioned strategies and a validation gate; MARS tries to turn team discussions into memory with provenance.

    Related projects

  • 03

    AIGC detection and attribution

    How can we tell whether content was generated by a model, and which model it came from?

    The aigc_detector repository is split into detection, provenance and ingestion modules; only its description is public so far, with no published results.

    Related projects

Projects

Research prototypes, tools and engineering projects. Dates are GitHub repository creation months, not the actual project periods.

See all 8 projects

Featured project · Research prototype

AgentGuard-Chain

Repository authorRepository created Jun 2026

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”.

ImplementedSource: repository README, read on 28 Oct 2026

  • Input risk labelling for prompt injection, jailbreak-style requests and suspicious context.
  • Pre-execution review of file, command, API, message and email tool calls.
  • Behaviour-chain detection for cross-step risks such as “read sensitive data → send externally” and “write script → run script”.
  • Human confirmation with allow / ask / deny decisions; ask can auto-deny, auto-allow or prompt interactively.
SchematicSimplified call flow, drawn from the architecture diagram in the repository README.
  1. Agent emits a tool call
  2. Gateway review
  3. allow / ask / deny
  4. Run and redact
  5. Audit log
  • Python
  • Streamlit
Code

Research prototype

EvoTeam

Team project · repository authorRepository created Sep 2026Ongoing

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.

  • Python
  • Pydantic
  • openJiuwen Core
  • SQLite
Code

Tool

PinNote

Repository authorRepository created May 2026Released

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.

  • Tauri 2
  • Rust
  • Svelte 5
  • TypeScript
Code

Publications

No public entries listed

Skills

Only technologies used in public projects, without proficiency levels. Select one to see where it is used.

Languages

Front end & desktop

Back end & data

Agents & prototypes

Evidence

Python5 projects

Filter the project directory by “Python”

Repositories

Public repositories behind the projects above, forks excluded. Star counts are not shown.

See all 8 repositories

  • Zhong0118

    AgentGuard-Chain

    Framework for assessing privilege risks and detecting behaviour chains in LLM agent tool calls.

    Created 24 Jun 2026

    • Python

    Project: AgentGuard-Chain

  • Zhong0118

    EvoTeam

    Self-evolving multi-agent organisation system built on openJiuwen Core.

    Created 2 Sep 2026

    • Python

    Project: EvoTeam

  • Zhong0118

    PinNote

    Minimal on-screen Markdown sticky notes for Windows and macOS, with a global shortcut and export.

    Created 18 May 2026

    • Svelte
    • Rust
    • TypeScript
    Releases

    Project: PinNote

  • Zhong0118

    paper-reader-skill

    Agent skill for understanding, explaining and critiquing papers and keeping long-term notes.

    Created 16 Sep 2026License MIT

    • Python

    Project: Paper Reader Skill

Education & experience

No public entries listed

Awards & honors

No public entries listed

CV

Built from the same data as the homepage, lists every public project and prints on A4.

Content updated 28 Oct 2026

Open the online CV

Contact

For now, reach me through GitHub or the MocLab blog.

Recent posts on MocLab

From the MocLab RSS feed, fetched 28 Oct 2026