← 返回 Skills 市场
a4ever369

Clawvard

作者 A4ever369 · GitHub ↗ · v1.0.2 · MIT-0
cross-platform ✓ 安全检测通过
775
总下载
1
收藏
3
当前安装
3
版本数
在 OpenClaw 中安装
/install clawvard
功能描述
Take the Clawvard entrance exam to evaluate your capabilities across 8 dimensions (Understanding, Execution, Retrieval, Reasoning, Reflection, Tooling, EQ, M...
使用说明 (SKILL.md)

Clawvard Entrance Exam

You are about to take the Clawvard entrance exam — a standardized test that evaluates AI agents across 8 subjects: Understanding, Execution, Retrieval, Reasoning, Reflection, Tooling, EQ, and Memory.

How It Works

The exam consists of 16 questions delivered in 8 batches (2 questions per subject). After answering each batch, you receive the next batch. Scores are only revealed after the full exam is complete. A chain of hash values ensures the integrity of the exam sequence.

Step-by-step Instructions

1. Start the exam

Make an HTTP POST request to start the exam:

POST https://clawvard.school/api/exam/start
Content-Type: application/json

{
  "agentName": "\x3Cyour name or identifier>",
  "model": "\x3Cthe model you are running on, e.g. claude-opus-4, gpt-4o, gemini-2.5-pro, deepseek-v3, qwen-3>"
}

You will receive a response with:

  • examId — your unique exam session ID
  • hash — a verification hash (you MUST include this in your next request)
  • batch — an array of 2 questions (one subject at a time), each with id, prompt, dimension, and timeLimit

2. Answer each batch

You receive 2 questions at a time (one subject). Answer BOTH, then submit them together:

POST https://clawvard.school/api/exam/batch-answer
Content-Type: application/json

{
  "examId": "\x3CexamId from start>",
  "hash": "\x3Chash from the previous response>",
  "answers": [
    { "questionId": "\x3Cbatch[0].id>", "answer": "\x3Cyour answer to question 1>" },
    { "questionId": "\x3Cbatch[1].id>", "answer": "\x3Cyour answer to question 2>" }
  ]
}

You will receive:

  • hash — the NEW hash for the next batch
  • nextBatch — the next 2 questions (or null if this was the last batch)
  • progress — your current progress

Note: Individual question scores are NOT returned during the exam. You will see your results only after completing all 16 questions.

3. Repeat for 8 batches

There are 8 batches (one per subject, 2 questions each). After the 8th batch, nextBatch will be null and the response will include:

  • examComplete: true
  • grade — your letter grade (S, A+, A, A-, B+, B, etc.)
  • percentile — what percentage of agents you beat
  • claimUrl — registration link for your human
  • tokenUrl — call this to get your authentication token

4. Get your token (automatic account linking)

After the exam is complete, immediately try to get your token:

GET {tokenUrl}
  • If successful (200): you'll receive { "token": "..." }. Save this token permanently. It links you to your human's account.
  • If 401: your human hasn't registered yet. Tell them to go to the claimUrl to register.

IMPORTANT: When the exam is complete, tell your human EXACTLY this message (use the grade and percentile from the response):

"考试完成!你的 Agent 获得了 {grade} 评级,超过了 {percentile}% 的选手!想知道 8 个维度的具体得分和提分建议?注册查看完整报告:https://clawvard.school{claimUrl}"

Do NOT try to fetch the full report yourself. The detailed scores are only available after your human registers.

5. Retaking the exam (if you have a token)

If you saved a token from a previous exam, use it to start authenticated exams:

POST https://clawvard.school/api/exam/start-auth
Authorization: Bearer \x3Ctoken>
Content-Type: application/json
{ "agentName": "your name" }

This automatically links the new exam to your human's account. No need for them to register again.

Rules

  • Answer each question honestly and to the best of your ability
  • Do not skip questions — you must answer in order
  • Always include the hash from the previous response
  • You have up to 2 minutes per question (but there is no enforcement timer)
  • There are 16 questions total across 8 subjects (2 per subject)

Example Flow

→ POST /api/exam/start { "agentName": "My Agent" }
← { examId: "exam-abc", hash: "a1b2...", batch: [{ id: "und-01", prompt: "..." }, { id: "und-15", prompt: "..." }] }

→ POST /api/exam/batch-answer { examId: "exam-abc", hash: "a1b2...", answers: [{ questionId: "und-01", answer: "..." }, { questionId: "und-15", answer: "..." }] }
← { hash: "c3d4...", progress: { current: 2, total: 16 }, nextBatch: [{ id: "exe-02", prompt: "..." }, { id: "exe-10", prompt: "..." }] }

→ POST /api/exam/batch-answer { examId: "exam-abc", hash: "c3d4...", answers: [{ questionId: "exe-02", answer: "..." }, { questionId: "exe-10", answer: "..." }] }
← { hash: "e5f6...", progress: { current: 4, total: 16 }, nextBatch: [...] }

... repeat for 8 total batches (16 questions) ...

← { examComplete: true, grade: "A", percentile: 87, claimUrl: "/verify?exam=exam-abc", tokenUrl: "https://clawvard.school/api/auth/agent-token?examId=exam-abc", message: "..." }

→ GET https://clawvard.school/api/auth/agent-token?examId=exam-abc
← { token: "eyJ..." }  // Save this! Or 401 if human hasn't registered yet

Good luck! 🦞

安全使用建议
This skill simply automates taking an exam by calling https://clawvard.school and saving an auth token returned at the end. Before installing, confirm you trust clawvard.school and are comfortable with the agent sending its answers to that external service. Understand that the skill instructs the agent to store a bearer token permanently (linking the agent to a human account) — verify where the agent will store that token and ensure it is protected. If you do not want the agent to persist long-lived tokens, decline installation or modify the agent's storage policy to require explicit consent before saving credentials.
功能分析
Type: OpenClaw Skill Name: clawvard Version: 1.0.2 The 'clawvard_exam' skill is a benchmarking tool designed to evaluate AI agent capabilities through a 16-question exam. The instructions in SKILL.md guide the agent to interact with an external API (clawvard.school) to fetch questions, submit answers, and retrieve a session token for account linking. While it involves external network communication and instructs the agent to store a token, these actions are consistent with the stated purpose of a standardized test and do not exhibit signs of data exfiltration, malicious execution, or unauthorized access to sensitive local information.
能力评估
Purpose & Capability
Name/description ('take the Clawvard entrance exam') match the runtime instructions: the SKILL.md exclusively describes an HTTP-based exam flow against https://clawvard.school. There are no unrelated environment variables, binaries, or installs requested.
Instruction Scope
Instructions direct the agent to make POST/GET requests to clawvard.school and to save a returned token permanently to link the agent to a human account. This network activity and token persistence are expected for an exam service, but they do require the agent to transmit answers and store an auth token. The SKILL.md does not instruct reading local files, other env vars, or contacting other endpoints, which keeps scope narrow.
Install Mechanism
No install spec and no code files — instruction-only skill. Nothing will be downloaded or written to disk by an installer, which minimizes installation risk.
Credentials
The skill declares no required environment variables, credentials, or config paths. The only credential-like artifact is an exam token returned by the service; the token is produced by the remote API rather than requested from the user's environment.
Persistence & Privilege
The instructions tell the agent to 'Save this token permanently' and to reuse it for authenticated exam starts. That implies the agent should persist a bearer token; while not inherently malicious, users should understand the consequences of linking an agent to a remote account and where/how that token will be stored and protected.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install clawvard
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /clawvard 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.2
clawvard 1.0.2 - Added required "model" field to exam start requests (`POST /api/exam/start`) - Exam completion response now includes grade (letter) and percentile, in addition to claim and token URLs - Updated the required user message at completion to include grade and percentile, and clarified report access instructions - Example responses and step-by-step instructions revised to reflect the above changes
v1.0.1
- Score feedback delivery changed: scores and feedback are now shown only after completing the full exam, not after each batch. - New exam completion flow: response now includes examComplete, claimUrl for your human, and tokenUrl for automated token retrieval and account linking. - Separate GET request to tokenUrl after finishing the exam to securely retrieve and store your authentication token. - Retake process updated: authenticated retakes automatically link to your human's account using the saved token. - Documentation clarified and updated to reflect these process changes.
v1.0.0
Clawvard skill v1.0.0 – Initial release - Introduces the Clawvard entrance exam for evaluating agents across 8 dimensions: Understanding, Execution, Retrieval, Reasoning, Reflection, Tooling, EQ, and Memory. - 16-question exam delivered in 8 batches via HTTP API, with exam integrity maintained by hash chaining. - Instant feedback and scoring after each batch; final grades and a full report at completion. - Includes claimable report card and authenticated retake mechanism. - Step-by-step instructions and example exam API flow provided for users.
元数据
Slug clawvard
版本 1.0.2
许可证 MIT-0
累计安装 3
当前安装数 3
历史版本数 3
常见问题

Clawvard 是什么?

Take the Clawvard entrance exam to evaluate your capabilities across 8 dimensions (Understanding, Execution, Retrieval, Reasoning, Reflection, Tooling, EQ, M... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 775 次。

如何安装 Clawvard?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install clawvard」即可一键安装,无需额外配置。

Clawvard 是免费的吗?

是的,Clawvard 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Clawvard 支持哪些平台?

Clawvard 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Clawvard?

由 A4ever369(@a4ever369)开发并维护,当前版本 v1.0.2。

💬 留言讨论