reveal-feedback（透露反馈）

Reveal Feedback Infrastructure

Reveal is a universal feedback platform where human reviewers screen-record themselves using products and provide AI-analyzed feedback. This skill connects to the Reveal REST API to manage the full feedback lifecycle — from collecting reviews to generating marketing content from that feedback.

Authentication

All API calls require the REVEAL_API_KEY environment variable. The key is a vendor API key generated from the Reveal dashboard under Settings → API Keys.

Every request uses this header:

Authorization: Bearer $REVEAL_API_KEY

Base URL: https://www.testreveal.ai/api/v1 (Override with REVEAL_BASE_URL env var if set.)

Capabilities

1. Check dashboard overview

Fetch products, active review tasks, and unread notifications to give the user a quick status update.

Steps:

1. GET /products to list vendor products
2. GET /review-tasks?status=active to list active tasks
3. GET /notifications?unread=true&limit=5 to get unread notifications
4. Summarize: product count, active tasks with submission progress, and recent notifications

2. Get feedback insights for a product

Fetch AI-aggregated insights: top issues, top positives, sentiment distribution, and suggestions.

Steps:

1. GET /products to find the product ID matching the user's request
2. GET /insights/{productId} to get aggregated insights
3. Present: sentiment breakdown, top issues ranked by frequency, top positives, unique issue count

3. Get product analytics

Fetch quantitative metrics for a product.

Steps:

1. GET /products/{productId}/analytics
2. Present: total submissions, analyzed count, average completion rate, sentiment distribution, top issues, top positives

4. View review submissions

Get individual review submissions with transcripts, AI analysis, sentiment, and issue counts.

Steps:

1. GET /review-tasks?status=active&limit=5 to find the relevant task (or use a task ID if provided)
2. GET /review-tasks/{taskId}/submissions to get all submissions
3. For each submission, present: status, sentiment, issue count, positive count, transcript preview

5. Create a review task

Create a new user-testing task so reviewers can test a product.

Steps:

1. GET /products to find the product matching the user's description
2. Extract from the user's message: title, objective, steps, feedback focus, reviewer count
3. POST /review-tasks with body:

{
  "title": "extracted title",
  "productId": "matched product ID",
  "requiredReviewers": 5,
  "instructions": {
    "objective": "what the reviewer should accomplish",
    "steps": "step-by-step instructions",
    "feedback": "what feedback to focus on"
  }
}

1. Confirm creation with task ID and details

6. Update a review task

Close, pause, or modify an existing review task.

Steps:

1. PATCH /review-tasks/{taskId} with fields to update (status, title, description, requiredReviewers)
2. Confirm the update

7. List products

Show all products registered on the vendor's Reveal account.

Steps:

1. GET /products?limit=50
2. Present each product: name, category, platform support (web/mobile), website

8. Get notifications

Check for new activity on Reveal.

Steps:

1. GET /notifications?unread=true&limit=20
2. Present notification messages with timestamps
3. If user says to mark as read: PATCH /notifications with {"markAllRead": true}

9. Register a webhook

Set up real-time event notifications.

Steps:

1. POST /webhooks with body:

{
  "url": "https://user-provided-url",
  "events": ["review.submitted", "review.analyzed", "task.completed", "video.generated"]
}

1. Return the webhook ID and signing secret. Instruct user to store the secret securely.

10. List webhooks

Steps:

1. GET /webhooks
2. Present each webhook: URL, subscribed events, active status

11. Summarize feedback into marketing highlights

Turn raw user feedback into polished, persuasive marketing statements.

Steps:

1. Collect the positive feedback strings (from insights or submissions)
2. POST /marketing/summarize-highlights with body:

{
  "highlights": ["feedback string 1", "feedback string 2", "..."]
}

1. Response contains summary — an array of marketing-ready highlight strings

12. Generate a marketing video script

Generate a 30-second video script from product feedback or the product description.

Steps:

1. GET /products to find the product ID
2. Optionally GET /insights/{productId} to extract positive highlights
3. POST /marketing/generate-script with body:

{
  "productId": "the product ID",
  "positiveHighlights": ["highlight 1", "highlight 2"],
  "scriptType": "problem-solution"
}

- positiveHighlights is optional; if omitted the product description is used - scriptType options: problem-solution, empathetic, aspirational, story-driven

1. Response contains script — a formatted script with [Scene X: Title] headers, Visual and Narration per scene

13. Generate a marketing image

Create a marketing image for a product with AI.

Steps:

1. GET /products to find the product ID
2. POST /marketing/generate-image with body:

{
  "productId": "the product ID",
  "prompt": "Modern dashboard showcasing the key analytics features",
  "styleName": "Modern Gradient",
  "tagline": "Insights that drive growth"
}

- prompt is required — the creative direction for the image - styleName and tagline are optional

1. Response contains imageUrl — a public URL to the generated image

14. Edit and refine a script

Parse a script into structured scenes and apply edits — change narration/visuals, add new scenes, or remove scenes.

Steps:

1. POST /marketing/edit-script with body:

{
  "script": "the current script text",
  "operations": [
    { "action": "edit", "sceneNumber": 3, "narration": "Updated narration text" },
    { "action": "add", "position": 4, "title": "New CTA", "visual": "Happy customer using phone", "narration": "Try it free today!" },
    { "action": "remove", "sceneNumber": 6 }
  ]
}

- operations is optional — omit it to just parse the script into structured scenes - action: "edit" — update title, visual, or narration of an existing scene (only include fields to change) - action: "add" — insert a new scene at position (or append if omitted) - action: "remove" — delete a scene by number (scenes renumber automatically)

1. Response contains both the updated script string and a scenes array with structured scene objects

15. Search stock media for video scenes

Find stock images and videos from the Pictory library to use in specific video scenes.

Steps:

1. GET /marketing/media-search?keyword=business+meeting&page=1

- keyword is required — the search term - category and page are optional

1. Response contains results — array of { url, type, thumbnail }
2. Use the url values in the customMediaUrls field of generate-video to assign media to specific scenes

16. Generate a marketing video (full pipeline)

Create a full marketing video from a script. This is a multi-step async workflow.

Steps:

1. Generate a script first (capability 12) or have the user provide one
2. Optionally refine the script (capability 14) — edit scenes, add/remove as needed
3. Optionally search for stock media (capability 15) to use in specific scenes
4. POST /marketing/generate-video with body:

{
  "productId": "the product ID",
  "script": "the full script with [Scene X: Title] headers",
  "voiceOver": { "speaker": "Jackson", "speed": 100 },
  "musicVolume": 0.4,
  "customMediaUrls": {
    "1": "https://stock-media-url-for-scene-1",
    "3": "https://stock-media-url-for-scene-3"
  }
}

- customMediaUrls is optional — map scene numbers to stock media URLs from media-search - Scenes without a custom URL will get auto-matched stock media based on the visual description Response: { "data": { "jobId": "...", "accessToken": "...", "productId": "..." } }

1. Wait ~30 seconds, then trigger rendering:

PUT /marketing/render-video/{jobId} with body:

{
  "accessToken": "the accessToken from step 4",
  "webhook": "https://optional-callback-url"
}

1. Poll for completion:

GET /marketing/video-status?jobId={jobId} Response: { "data": { "status": "in-progress" | "completed" | "failed", "videoUrl": "..." } } - Keep polling every 15-30 seconds until status is completed - When complete, videoUrl contains the final video URL

Response format

All API responses follow this structure:

• Success: { "data": { ... } }
• Error: { "error": { "code": "ERROR_CODE", "message": "description" } }

Guardrails

• Never expose or log the API key in responses to the user
• If an API call fails with 401, tell the user their API key may be invalid or expired
• If a product is not found, suggest listing products first
• When creating review tasks, always confirm the details with the user before sending the POST
• Do not fabricate feedback data — only report what the API returns
• Marketing endpoints require a vendor API key — reviewer keys will get a 403
• Video generation is async — always poll video-status rather than assuming immediate completion
• When generating videos, save the accessToken from generate-video — it's needed for render-video