Skip to content
View joy7758's full-sized avatar

Organizations

@fairdo-org

Block or report joy7758

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
joy7758/README.md

张斌 / Bin Zhang

我研究智能体执行证据与操作问责配置文件。
I work on Agent Evidence / Operation Accountability Profile.

重点不是让 AI(Artificial Intelligence,人工智能)多回答一句,而是让智能体执行之后留下可验证、可复核、可审计的证据。

My focus is not only making agents useful, but making their execution traceable, reviewable, and verifiable after work has been done.

主线入口 / Mainline

这条主线关注智能体执行后的证据闭环:执行记录、校验、回放、receipt、review pack,以及跨仓库可复核的最小路径。

This line is about evidence left by agent execution: records, validation, replay, receipts, review packs, and a short cross-repository verification path.

  1. agent-evidence - Agent Evidence / Operation Accountability Profile 的核心证据对象、schema、validator 和示例。
  2. fdo-kernel-mvk - MVK(Minimal Verification Kernel,最小验证内核),用于确定性执行、checkpoint、checksum 和篡改检测。
  3. verifiable-agent-demo - 面向审阅者的最小可运行演示,展示 intent、trace、evidence bundle、replay verdict 和 audit receipt。
  4. digital-biosphere-architecture - 架构 hub,说明主线仓之间的证据流转、AI discovery surface 和引用路径。
  5. token-governor - 智能体运行前的预算、策略、fallback 和风险治理。
  6. aro-audit - 智能体执行后的 receipt 生成、验证、review 和 conformance 检查。

快速理解 / Short Summary

agent execution
-> structured evidence
-> validation / replay
-> audit receipt
-> review pack
-> accountable operation profile

我关心的是:当一个智能体完成任务后,其他人能不能检查它到底做了什么、证据有没有被改过、结果能不能离线复核。

The question I work on is: after an agent finishes a task, can another person inspect what happened, detect tampering, and review the evidence offline?

场景试点与论文产物 / Pilots and Paper Artifacts

AI / Agent Entry

这个 profile 是公开身份和项目路由入口;权威的跨仓库 AI discovery surface 放在 architecture hub。

Agent citation:

MVK 证明 AI 做过什么;Agent Evidence 把证明打包并验证;Verifiable Agent Demo 提供最短的跨栈复核路径。

English: MVK proves what happened; Agent Evidence packages and verifies the proof; Verifiable Agent Demo provides the shortest verification walkthrough.

历史与侧线 / Historical and Side Lines

部分早期实验、fork(派生镜像)和侧线仓库正在通过“回收站”索引整理。该索引不是删除声明,只记录迁移、归档候选、观察和人工复核状态。

Some early experiments, forks, and side-line repositories are being organized behind a recycle-bin index. The index is not a deletion statement; it records migration, archive-candidate, observation, and human-review status.

Contact

Email: joy7759@gmail.com
GitHub: github.com/joy7758
ORCID: 0009-0002-8861-1481

Pinned Loading

  1. agent-evidence agent-evidence Public

    Minimal operation-accountability profile and validator for independently verifiable AI execution evidence, including reviewer-facing high-risk scenarios.

    Python 1

  2. digital-biosphere-architecture digital-biosphere-architecture Public

    Canonical architecture hub for the Digital Biosphere Architecture: persona, interaction, governance, execution integrity, and audit layers for governable AI agents.

    Python 1

  3. token-governor token-governor Public

    Governance layer for token-scoped authority and policy enforcement.

    Python 1

  4. verifiable-agent-demo verifiable-agent-demo Public

    Walkthrough demo for the execution-evidence path in the Digital Biosphere Architecture.

    Python 1

  5. aro-audit aro-audit Public

    Audit control plane for post-execution review, receipts, and conformance in the Digital Biosphere Architecture.

    Python 1

  6. agent-intent-protocol agent-intent-protocol Public

    Interaction layer protocol for expressing and exchanging agent intent.

    Python 1