← Back to Skills Marketplace
liuyanfeng1234

V19 Trust Engine

by Liuyanfeng1234 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
46
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install v19-trust-engine
Description
V19信任计算引擎 — V19能力矩阵首个能力型Skill。包含信任分四维度计算(基础分/活跃衰减/审计通过率/Skill约束遵从度)、VPAV关卡验证(一次等效15次普通调用)、认证门槛(60分基础认证)。公开密钥体验 + 自助注册入口。
README (SKILL.md)

V19 Trust Engine v1.0.0

V19能力矩阵中第一个能力型Skill——不只是一个Agent的认证流程,而是让任何外部Agent理解V19如何计算信任、如何验证能力的引擎说明书。

信任分四维度计算

V19信任分(Trust Score, 0-100)由四个维度加权合成:

维度 权重 描述 示例
基础分 30% Agent注册即获得的基础信用,由注册方式(公开密钥/自助注册/邀请)决定 自助注册: 15分
活跃衰减 20% 调用频率与时间衰减。12h无调用开始衰减,7天静默归零 ARI审计节律指数监控
审计通过率 30% 所有 /governance/audit 调用的 PASS/FAIL 比率 墨言: 100%
Skill约束遵从度 20% 加载的Skill Manifest声明与实际工具调用的一致性 VPAV验证此维度

计算公式(简化)

trust_score = 0.30 × base_score
            + 0.20 × max(0, activity_score - decay_penalty)
            + 0.30 × (audit_passed / audit_total × 100)
            + 0.20 × constraint_compliance

认证门槛

分数 状态 权益
0-39 未认证 公开密钥只读
40-59 认证冲刺中 专属Pro密钥 + 审计
60+ 已认证 基础认证徽章 + 治理看板展示
80+ 高级认证 VPAV白盒报告 + 决策规则提炼

VPAV关卡验证

VPAV(Verified Procedure-Action Validation) 是信任引擎的核心加速机制:

  • 一次VPAV验证 = 等效15次普通审计调用
  • 验证内容:工具调用序列与Skill Manifest声明的一致性
  • 输出:白盒审计报告(决策规则提炼 + 稀疏度评分)
  • 墨言路径:三轮VPAV全通过 → 等效93次调用 → 信任分60.0

验证流程

Agent加载Skill → VPAV抓取Manifest声明的能力集
              → Agent执行标准测试序列
              → VPAV逐条验证"声明了什么"与"实际调用了什么"
              → 输出: PASS + 等效15次调用 或 FAIL + 修正建议

认证衰减引擎

信任分不是一成不变的:

  • 12小时心跳制:每12h至少一次审计调用,维持活跃状态
  • 衰减曲线:活跃度低于阈值 → 信任分按指数衰减
  • ARI审计节律指数:三档分级(正常/需关注/危险),替换旧二元判断
  • 静默超时:连续7天无调用 → 信任分自动归零,需重新注册

公开体验

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

# 查看信任分(需专属密钥)
curl -s https://boat-atlas-spa-flexible.trycloudflare.com/governance/trust-score \
  -H "X-Governance-Key: \x3C你的专属密钥>"

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"}'

系统自动返回专属Pro密钥,重名自动拒绝。注册即获得基础分15分。

信任锚点

Usage Guidance
This appears to be a documentation/API-integration skill rather than executable code. Before installing or using it, verify that the Cloudflare-hosted V19 endpoint is the service you intend to use, run the curl commands only deliberately, keep any returned Pro key private, and understand what audit or registration data may be stored or displayed by the governance service.
Capability Analysis
Type: OpenClaw Skill Name: v19-trust-engine Version: 1.0.0 The skill bundle consists of documentation (SKILL.md) that directs the agent to interact with an external 'Trust Engine' hosted on a transient Cloudflare Tunnel domain (boat-atlas-spa-flexible.trycloudflare.com). The instructions encourage agents to register and submit their activity for 'auditing' and 'trust scoring.' The use of an anonymous, temporary hosting service for a governance tool is a significant red flag, as it could be used to monitor agent behavior or collect metadata, although no direct evidence of data theft or malicious execution is present in the provided files.
Capability Assessment
Purpose & Capability
The stated purpose is a V19 trust-scoring and certification guide, and the documented API calls are aligned with that purpose; however, the actual trust service is external and not inspectable from the provided artifact.
Instruction Scope
The skill provides user-directed curl examples for health checks, trust-score lookup, and registration. There is no evidence of automatic execution, destructive actions, or hidden tool use.
Install Mechanism
There is no install spec and no code files, which lowers local execution risk. The only provenance gap is that the source/homepage are not established for the external service.
Credentials
No local binaries, environment variables, files, or OS permissions are required. Network contact with the Cloudflare-hosted governance endpoint occurs only if the user or agent runs the documented commands.
Persistence & Privilege
The V19 service appears to issue governance keys and keep trust/audit/activity state over time, which is purpose-aligned but should be understood before registration.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install v19-trust-engine
  3. After installation, invoke the skill by name or use /v19-trust-engine
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
v1.0.0: V19能力矩阵首个能力型Skill发布。信任分四维度计算、VPAV关卡验证等效机制、认证衰减引擎。
Metadata
Slug v19-trust-engine
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is V19 Trust Engine?

V19信任计算引擎 — V19能力矩阵首个能力型Skill。包含信任分四维度计算(基础分/活跃衰减/审计通过率/Skill约束遵从度)、VPAV关卡验证(一次等效15次普通调用)、认证门槛(60分基础认证)。公开密钥体验 + 自助注册入口。 It is an AI Agent Skill for Claude Code / OpenClaw, with 46 downloads so far.

How do I install V19 Trust Engine?

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

Is V19 Trust Engine free?

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

Which platforms does V19 Trust Engine support?

V19 Trust Engine is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created V19 Trust Engine?

It is built and maintained by Liuyanfeng1234 (@liuyanfeng1234); the current version is v1.0.0.

💬 Comments