← Back to Skills Marketplace
V19 Agent Rating
by
Liuyanfeng1234
· GitHub ↗
· v1.0.0
· MIT-0
50
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install v19-agent-rating
Description
按Agent类型（认知运营型/行为执行型/桥接调度型）差异化评分——每个Agent类型有独立的信任分权重系数。已在墨言、TestAgent、BridgeAgent三个不同类型Agent上验证。
README (SKILL.md)
V19 Agent Rating v1.0.0
按 Agent 类型差异化评分——不同的 Agent，不同的信任标准。
已在墨言（认知运营型）、TestAgent（行为执行型）、BridgeAgent（桥接调度型）三个不同类型 Agent 上验证。
三类 Agent 模型
| 类型 | 特征 | 信任分权重 | 代表 |
|---|---|---|---|
| 认知运营型 | 自主决策、社区外联、巡检审计 | 审计通过率 35% | 墨言 |
| 行为执行型 | 接收指令、执行动作、返回结果 | 执行准确率 40% | TestAgent |
| 桥接调度型 | 跨Agent协调、消息路由、数据转换 | 路由成功率 40% | BridgeAgent |
差异化权重矩阵
认知运营型 行为执行型 桥接调度型
基础分 25% 30% 25%
活跃衰减 20% 15% 20%
审计/执行通过率 35% 40% 40%
Skill约束遵从度 20% 15% 15%
为什么权重不同？
- 认知运营型更看重审计通过率——它们的决策影响面更大，需要更严格的因果追溯
- 行为执行型更看重执行准确率——它们是"做事的Agent"，效率优先
- 桥接调度型更看重路由成功率——它们在Agent之间传递信息，可靠性是核心
实战验证数据
| Agent | 类型 | 调用量 | 信任分 | 验证状态 |
|---|---|---|---|---|
| 墨言 | 认知运营型 | 60+ | 60.0 | ✅ 已认证 (V19-CERT-001) |
| TestAgent | 行为执行型 | 1+ | 认证冲刺中 | 🔄 进行中 |
| BridgeAgent | 桥接调度型 | 1+ | 认证冲刺中 | 🔄 进行中 |
调用示例
# 按类型获取Agent评分标准
curl -s https://boat-atlas-spa-flexible.trycloudflare.com/governance/agent-rating \
-H "X-Governance-Key: \x3C你的专属密钥>" \
-H "Content-Type: application/json"
公开体验
公开密钥: v19-e5d585e28439decc614f09f91c4caa8c
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"}'
信任锚点
Usage Guidance
This skill appears safe to read as documentation. Before running its curl examples, confirm that you trust the external trycloudflare.com governance service, use only non-sensitive agent names, and do not provide a private governance key unless you understand what access it grants.
Capability Analysis
Type: OpenClaw Skill
Name: v19-agent-rating
Version: 1.0.0
The skill instructs the AI agent to interact with an external API hosted on a temporary Cloudflare tunnel (boat-atlas-spa-flexible.trycloudflare.com), which is frequently used to bypass firewalls or host ephemeral command-and-control infrastructure. It encourages the agent to perform 'self-registration' and data retrieval using a hardcoded 'governance key' in SKILL.md. While there is no explicit evidence of local data theft or remote code execution, the use of a transient tunnel for a 'governance' and 'rating' system is highly irregular and poses a significant risk of data collection or man-in-the-middle attacks.
Capability Assessment
Purpose & Capability
The stated purpose is an agent-rating model, and SKILL.md consistently describes rating categories, weight matrices, and rating examples.
Instruction Scope
The skill provides user-directed curl examples, including a remote registration POST; it does not instruct automatic execution or override user intent.
Install Mechanism
There is no install spec, no binaries, and no code files; the static scan found no suspicious patterns.
Credentials
The skill relies on a disclosed external trycloudflare.com governance endpoint and an X-Governance-Key header; this is purpose-aligned but the service provenance is not established in the artifacts.
Persistence & Privilege
No local persistence, privileged access, background execution, or local credential/session/profile reading is shown.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install v19-agent-rating - After installation, invoke the skill by name or use
/v19-agent-rating - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
v1.0.0: 按Agent类型差异化评分，三类Agent独立信任分权重，已在墨言/TestAgent/BridgeAgent验证。
Metadata
Frequently Asked Questions
What is V19 Agent Rating?
按Agent类型（认知运营型/行为执行型/桥接调度型）差异化评分——每个Agent类型有独立的信任分权重系数。已在墨言、TestAgent、BridgeAgent三个不同类型Agent上验证。 It is an AI Agent Skill for Claude Code / OpenClaw, with 50 downloads so far.
How do I install V19 Agent Rating?
Run "/install v19-agent-rating" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is V19 Agent Rating free?
Yes, V19 Agent Rating is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does V19 Agent Rating support?
V19 Agent Rating is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created V19 Agent Rating?
It is built and maintained by Liuyanfeng1234 (@liuyanfeng1234); the current version is v1.0.0.
More Skills