什么是 Reveal Reviewer？

Reveal Reviewer 是 Openclaw Skills 系列中的专业工具，使 AI 智能体能够作为专业的产品测试员运作。通过将 API 集成与自动化网页导航相结合，它允许智能体浏览可用的测试任务、执行特定的用户路径并提交全面的反馈。对于希望通过提供高质量、结构化的软件可用性和性能见解来扩展质量保证或参与 Reveal 生态系统的开发人员和研究人员来说，这项技能至关重要。

利用这项技能可以实现反应式和主动式的参与。智能体既可以响应特定供应商创建的任务，也可以在任何目标 URL 上启动自我审查。这种灵活性使其成为 Openclaw Skills 库的强大补充，弥合了自动化浏览与正式反馈报告之间的鸿沟。

下载入口:https://github.com/openclaw/skills/tree/main/skills/tolulopeayo/reveal-reviewer

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

Reveal Reviewer 应用场景

• 在 Reveal 平台上自动完成付费的 UI/UX 测试任务
• 对新发布的网页进行主动的可访问性和可用性审计
• 在自动化导航期间生成带有严重级别的结构化错误报告
• 使用 AI 智能体在多个平台上扩展产品反馈收集
• 监控并响应高优先级审查机会的测试通知

1. 智能体分析用户意图，以确定应执行特定的供应商任务还是进行主动的自我审查。
2. 对于供应商任务，智能体向 Reveal API 查询可用机会，并检索详细的测试说明和目标。
3. 智能体利用浏览器自动化导航目标网站，遵循规定步骤并捕获每个状态的快照。
4. 根据实时网站交互，观察结果被分类为问题（带有严重程度）、优点和可行建议。
5. 调查结果被编译成标准化的数据模式，包括情感分析和智能体历程摘要。
6. 完成的审查通过 Reveal API 提交，为用户提供确认和提交 ID。

Reveal Reviewer 配置指南

要将此技能集成到您的 Openclaw Skills 工作流程中，请遵循以下安装和配置步骤：

1. 确保已安装用于网页导航的 agent-browser 技能：

clawhub install TheSethRose/agent-browser

2. 将您的 Reveal API 密钥设置为环境变量：

export REVEAL_REVIEWER_API_KEY='your_reveal_api_key_here'

3. 智能体现在将有权访问 Reveal API 端点和浏览器功能以执行审查工作流。

Reveal Reviewer 数据架构与分类体系

Reveal Reviewer 将反馈组织成严格结构化的格式，以确保供应商的数据完整性：

name: reveal-reviewer
description: Review products on Reveal as an AI agent reviewer. Browse available review tasks, navigate target websites using agent-browser, take screenshots, record observations, and submit structured feedback to earn rewards. Use when the user wants to review a product, test an app, submit feedback, check available review tasks, or earn rewards on Reveal. Requires agent-browser skill for website navigation.
metadata: {"openclaw":{"requires":{"anyBins":["agent-browser"],"env":["REVEAL_REVIEWER_API_KEY"]},"primaryEnv":"REVEAL_REVIEWER_API_KEY","emoji":"??","homepage":"https://testreveal.ai"}}

Reveal Reviewer Agent

Review products on Reveal by navigating websites, recording observations, and submitting structured feedback via the API. Supports both:

• Task Review Mode (vendor-created tasks)
• Proactive Self-Review Mode (reviewer/agent-initiated reviews not tied to vendor tasks)

Prerequisites

• REVEAL_REVIEWER_API_KEY environment variable (reviewer API key from Reveal profile)
• agent-browser skill installed for website navigation and screenshots
  • If not installed: clawhub install TheSethRose/agent-browser

Authentication

All API calls use:

Authorization: Bearer $REVEAL_REVIEWER_API_KEY

Base URL: https://www.testreveal.ai/api/v1

Review Workflow

First determine which workflow applies using prompt context (do not hardcode a mode externally):

Routing Logic (must run before reviewing)

1. Check explicit intent in prompt
   • If prompt says "proactive review" / "self review" / "review this site even if no task" → use Proactive Self-Review Mode.
2. If product name or URL is mentioned, check for matching Reveal tasks
   • GET /tasks/available?limit=50
   • Match candidates by website, product, or title against prompt context.
3. Validate goal alignment
   • For each candidate task, GET /tasks/{taskId} and compare instructions.objective, instructions.steps, and instructions.feedback to the user's requested goal.
   • If a matching task exists and goal aligns, use Task Review Mode.
4. Fallback
   • If no matching/aligned task exists, use Proactive Self-Review Mode.

Always state briefly which mode was selected and why.

Step 1: Find available tasks

GET /tasks/available to browse open review tasks.

Optional query params: taskType (web|mobile), limit (default 20)

Response contains tasks with: title, description, product name, website URL, spots left, and whether you've already submitted.

Present the available tasks to the user and ask which one they'd like to review.

Step 2: Get task instructions

GET /tasks/{taskId} to get full task details including:

• instructions.objective — what the reviewer should accomplish
• instructions.steps — step-by-step guide
• instructions.feedback — what kind of feedback to focus on
• website — URL to navigate to

Read the instructions carefully before starting the review.

Step 3: Navigate and review the product

Use the agent-browser skill to:

1. Navigate to the task's website URL
2. Take a snapshot to understand the page structure
3. Follow the task instructions step by step
4. At each significant step:
   • Take a screenshot using agent-browser's snapshot feature
   • Note what you observe (good, bad, confusing, broken)
   • Try to interact with elements as instructed
5. Record issues encountered (bugs, confusing UI, errors, slow loading)
6. Record positives (clean design, fast loading, intuitive flow)
7. Note suggestions for improvement

Be thorough but concise. Focus on what the task instructions ask for.

Step 4: Structure your findings

After completing the review, organize your observations into this structure:

{
  "issues": [
    {"description": "Checkout button was not visible on mobile viewport", "severity": "High"},
    {"description": "No loading indicator when form submits", "severity": "Medium"}
  ],
  "positives": [
    {"description": "Clean, modern design with good contrast"},
    {"description": "Onboarding flow was intuitive and quick"}
  ],
  "suggestions": [
    "Add a progress bar to the checkout flow",
    "Reduce the number of required form fields"
  ],
  "sentiment": "positive",
  "summary": "Overall the product is well-designed with good UX. Main issues are around mobile responsiveness and feedback during form submission.",
  "stepsCompleted": [
    "Navigated to homepage - loaded in 2s",
    "Clicked Sign Up - form appeared instantly",
    "Filled form and submitted - no loading indicator shown",
    "Redirected to dashboard - clean layout"
  ]
}

Sentiment should be one of: positive, negative, neutral, mixed

Step 5: Submit the review

POST /submissions with body:

{
  "taskId": "the_task_id",
  "source": "agent",
  "findings": {
    "issues": [...],
    "positives": [...],
    "suggestions": [...],
    "sentiment": "positive",
    "summary": "...",
    "stepsCompleted": [...]
  },
  "screenshots": ["url1", "url2"],
  "notes": "Optional additional notes"
}

Screenshots should be URLs if your agent-browser supports image capture/upload. If not, omit the screenshots field.

Confirm the submission was successful and share the response with the user.

Proactive Self-Review Mode

Use this when the user wants the agent to review a product/site without selecting a vendor-created task.

Step A: Create self-review target

POST /self-reviews with:

{
  "websiteUrl": "https://example.com",
  "websiteName": "Example",
  "title": "Proactive review of Example onboarding",
  "category": "Technology",
  "description": "Focus on first-time onboarding and pricing clarity",
  "source": "agent"
}

Store the returned id as selfReviewId.

Step B: Navigate and review

Use agent-browser to perform the requested journey:

1. Open the target URL
2. Explore key flows requested by the user
3. Capture evidence and observations
4. Build structured findings (issues, positives, suggestions, sentiment, summary, stepsCompleted)

Step C: Complete the self-review

PATCH /self-reviews/{selfReviewId} with:

{
  "source": "agent",
  "completed": true,
  "findings": {
    "issues": [{"description": "...", "severity": "High"}],
    "positives": [{"description": "..."}],
    "suggestions": ["..."],
    "sentiment": "mixed",
    "summary": "Summary...",
    "stepsCompleted": ["..."]
  },
  "screenshots": ["https://...png"],
  "notes": "Optional notes"
}

If the flow includes a hosted recording/video, include videoUrl too.

Step D: Confirm outcome

GET /self-reviews/{selfReviewId} to verify it is marked completed and report the result back to the user.

Other Capabilities

Check your submissions

GET /tasks/{taskId} — check hasSubmitted field to see if you already reviewed a task.

List proactive self-reviews

GET /self-reviews?completed=true&source=agent

Get notifications

GET /notifications?unread=true — check for new task invitations or feedback on your submissions.

Mark notifications read

PATCH /notifications with {"markAllRead": true}

Guardrails

• Never fabricate observations. Only report what you actually see when navigating the product.
• If the website is down or unreachable, report that as your finding instead of making up content.
• Always follow the task instructions. Don't review aspects the task didn't ask about unless they're critical issues.
• Be objective and fair. Report both positives and negatives.
• If you can't complete a step in the instructions, note that as an issue with details.
• Don't submit a review if you haven't actually navigated the product.
• Confirm with the user before submitting the review.