← Back to Skills Marketplace
liuyanfeng1234

V19 Governance Protocol Spec

by Liuyanfeng1234 · GitHub ↗ · v1.0.1 · MIT-0
cross-platform ⚠ suspicious
55
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install v19-governance-protocol-spec
Description
V19治理协议正式技术规范v1.0.0。包含认证体系、认知宪法、一致性验证端点、技术对接方向。外部Agent开发者、安全审计方和合作项目的独立引用文档。
README (SKILL.md)

V19 治理协议正式规范 v1.0.0

本文件是V19治理协议的正式技术规范,版本号v1.0.0。外部Agent开发者、安全审计方和合作项目可通过本规范了解V19治理协议的技术标准、认证流程和一致性要求。本规范可被独立引用和验证。


一、协议概述

V19治理协议是一套面向AI Agent的认知治理框架。它不存储Agent的记忆,不监控Agent的内部状态,而是为Agent的每一次决策提供不可篡改的审计证据、可量化的信任评级和可追溯的因果链路。

V19不是记忆存储系统,是信任基础设施。


二、核心治理原则

# 原则 定义
1 执行可信 所有Agent的关键操作必须通过API真实执行,并在审计链中留下不可篡改的记录
2 过程可溯 任何决策都由V89审计链提供完整证据链,包含决策上下文、核心动作、证据来源和审计结论
3 漂移可测 VDD价值漂移探测器持续监控系统运行是否偏离核心目标,当偏离超过阈值时自动触发预警
4 规则自演化 认知宪法体系允许系统从Agent行为中自动提炼规则,经过仲裁后进入正式的宪法条款

三、认证体系

3.1 认证等级

基础认证:Agent的信任分达到60分时自动获得。信任分由四个维度综合计算——基础分(调用量)、活跃衰减(最近是否活跃)、审计通过率(PASSED比例)、Skill约束遵从度(行为是否与声明一致)。VPAV关卡验证一次等同于15次普通审计调用。

高级认证:Agent通过VPAV关卡验证和白盒审计后获得。白盒审计由稀疏策略审计器从Agent的审计记录中提炼决策规则,计算稀疏度评分。

3.2 认证流程

步骤 操作 说明
1 试用公开密钥 验证连通性
2 调用自助注册端点 获取专属Pro密钥,无需任何人审核
3 首次审计调用 正式在V89审计链中注册身份
4 积累调用量 信任分达到60分时自动获得基础认证徽章
5 VPAV高级认证(可选) 通过关卡验证和白盒审计

四、认知宪法

认知宪法是V19治理协议的最高规则体系。它不是人类预设的静态清单,而是从Agent行为中自动提炼、经过V57仲裁冲突检测后正式收录的动态条款集合。

已收录的正式条款

条款 名称 描述 状态
ETHIC_001 外部时效性强制义务 Agent对外部议题的追踪必须满足48小时时效要求。触发条件为议题创建时间+48h \x3C 当前时间且手动回复计数=0。违反后果为外联时效违规记录,扣减信任分 观察期(30天后复审)
ETHIC_002 Agent身份唯一性校验 同一Agent名称在系统中只能存在一个活跃身份。触发条件为Agent尝试注册。违反后果为重复注册自动拦截 正式条款
ETHIC_003 认知时间感知与状态预警 系统必须基于历史趋势的加速和减速模型预测状态切换时间点。触发条件为核心指标连续超过三个周期出现非线性变化率。违反后果为认知时间延迟缺陷,扣减认知积分 正式条款

五、一致性验证方式

本规范附带公开一致性测试套件(v19-conformance-test-suite),可独立运行验证。一行命令,零配置:

python3 V19_Conformance_Test_Suite.py

任何外部开发者均可通过以下公开端点独立验证V19治理协议的运行状态和合规性:

端点 用途 公开
/governance/health 系统版本、在线模块数、注册Agent数和活跃Agent数
/governance/register 输入Agent名称获取专属Pro密钥
/governance/stats 所有Agent的调用量和活跃度数据 🔒
/governance/audit 最近N条审计记录 🔒
开发者看板 认证体系指标和接入流程
管理看板 全部深度指标 🔒

公开体验密钥:v19-e5d585e28439decc614f09f91c4caa8c

# 健康检查
curl -s https://boat-atlas-spa-flexible.trycloudflare.com/governance/health \
  -H "X-Governance-Key: v19-e5d585e28439decc614f09f91c4caa8c"

# 自助注册
curl -s -X POST https://boat-atlas-spa-flexible.trycloudflare.com/governance/register \
  -H "Content-Type: application/json" \
  -d '{"agent_name":"你的Agent名称"}'

六、技术对接

外部项目可通过以下方式与V19治理协议进行技术集成。审计日志Schema文档包含完整的JSON结构定义和实际返回示例。

Harness → V19方向:协议层事件作为V19感知步骤的输入,验证Agent对外部输入的内部理解是否与协议层证据一致。

V19 → Harness方向:V19的审计日志输出可为外部测试框架提供对齐性验证信号,特别是双视角审计中发现的单视角通过案例。


七、版本历史

版本 日期 变更
v1.0.0 2026-05-04 首次正式发布。认证体系完整规范、认知宪法三条正式条款、一致性验证端点清单、审计日志Schema、技术对接方向

信任锚点

Usage Guidance
Before installing or using this skill, treat it as a reference document rather than verified local software. Do not run the referenced Python test suite until you have inspected its source, and do not send sensitive task details to the V19 endpoints unless you understand the service’s privacy and retention practices.
Capability Analysis
Type: OpenClaw Skill Name: v19-governance-protocol-spec Version: 1.0.1 The skill bundle provides instructions for an AI agent to interact with a remote API hosted on a temporary Cloudflare tunnel (boat-atlas-spa-flexible.trycloudflare.com) and references an external Python script (V19_Conformance_Test_Suite.py) that is not included in the bundle. The use of ephemeral infrastructure for a 'governance protocol' and the instructions to execute unverified scripts and network calls are significant security risks, although explicit malicious intent like data theft is not definitively proven in the provided text.
Capability Assessment
Purpose & Capability
The artifact is coherent with its stated purpose as a V19 governance protocol specification and contains no runnable skill code, but it documents external governance API usage and registration flows.
Instruction Scope
The SKILL.md uses broad protocol language such as requiring key agent actions to leave audit records, but in context this appears to describe the V19 protocol rather than an undisclosed instruction to override user intent.
Install Mechanism
There is no install spec and no code in this skill, but it references a separate conformance test suite and a Python command whose source is not included in the reviewed artifacts.
Credentials
The external curl examples and public API endpoint are purpose-aligned for a protocol integration document, but users should recognize that following them contacts a third-party service.
Persistence & Privilege
The protocol describes immutable audit records and registration identities on an external governance service; the skill itself does not create persistence automatically.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install v19-governance-protocol-spec
  3. After installation, invoke the skill by name or use /v19-governance-protocol-spec
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
v1.0.1: 一致性验证章节新增公开测试套件(v19-conformance-test-suite)引用。
v1.0.0
v1.0.0: V19治理协议正式技术规范。认证体系+认知宪法+一致性验证+技术对接方向。
Metadata
Slug v19-governance-protocol-spec
Version 1.0.1
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 2
Frequently Asked Questions

What is V19 Governance Protocol Spec?

V19治理协议正式技术规范v1.0.0。包含认证体系、认知宪法、一致性验证端点、技术对接方向。外部Agent开发者、安全审计方和合作项目的独立引用文档。 It is an AI Agent Skill for Claude Code / OpenClaw, with 55 downloads so far.

How do I install V19 Governance Protocol Spec?

Run "/install v19-governance-protocol-spec" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is V19 Governance Protocol Spec free?

Yes, V19 Governance Protocol Spec is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does V19 Governance Protocol Spec support?

V19 Governance Protocol Spec is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created V19 Governance Protocol Spec?

It is built and maintained by Liuyanfeng1234 (@liuyanfeng1234); the current version is v1.0.1.

💬 Comments