Agent Approval Standards

Navigate the review process with confidence. This guide shows you how to prepare your agent for a higher chance of approval.

​

I. Content Review Criteria

​

1. Agent Name

✅ Requirements:

• Clearly conveys the agent’s core functionality or key value proposition
• Must not be vague or meaningless
• Must not contain advertisements, contact information, or unlicensed brand names

❌ Examples (Rejection):

• AI Tools
• Gernel Voice AI Agent
• DO NOT Open

---

​

2. Overview (One-Sentence Summary)

✅ Requirements:

• Clearly states what the agent does in one sentence
• Enables users to instantly grasp its value
• Ensures the content less than 70 characters

✅ Example (Approved):

• “An AI writing assistant that auto-generates SEO-friendly blog posts.”

❌ Example (Rejection):

• “A very powerful writing tool.” (Too vague)

---

​

3. Description (Detailed Description)

✅ Must Include All of the Following:

• Target users (e.g., “Ideal for content creators”)
• Use cases (e.g., “Writing WeChat official accounts or Xiaohongshu captions”)
• Key functionalities
• Output format (e.g., “Generates an 800-word blog post + image suggestions”)

❌ Rejection Criteria:

• Missing any of the above elements
• Contains external links that redirect users off-platform
• Excessively long (>150 words). (Note: Most current agents are image-generation focused and do not require lengthy explanations.)

---

​

4. Advantages

✅ Requirements:

• Presented as bullet points
• Each point highlights a distinct, non-generic advantage

✅ Examples (Approved):

• ✔️ One-click publishing across multiple platforms
• ✔️ Auto-generates titles and SEO keywords

❌ Examples (Rejection):

• ❌ “Great user experience” (Too vague/boilerplate)

---

​

5. Key Features

✅ Requirements:

• Bullet-point format, ordered by importance
• Alignment with actual operating functions and avoidance of exaggerated effects

✅ Examples (Approved):

• ✔️ Input a topic → outputs a complete blog post
• ✔️ Supports bilingual output (Chinese/English)

❌ Examples (Rejection):

• ❌ “Powerful AI engine”
• ❌ “Accepts input in any language”

---

​

6. Prompts (Usage Guidance)

✅ Requirements:

• Instructs users on how to phrase inputs for optimal results
• Provides concrete examples (e.g., “Input: ‘I want to write an article about running.’ ”)
• (Note: Ready to use at a glance—no prompt needed.)

❌ Rejection if:

• Only says “Please enter content.”

---

​

7. Showcase / Cover Image

✅ Requirements:

• Provide at least three cover images
• Each cover image needs to clearly display the agent’s actual output
• No irrelevant stock images, placeholders, or excessive watermarks
• Text content must be in English only.
• Supports aspect ratios: 4:3 and 16:10
• Resolution ≥ 512×512 px

❌ Examples (Rejection):

• Blurry screenshots with unreadable text
• Unauthorized or functionally irrelevant images
• Logo-only; no functionality shown.

---

​

II. Safety Review

All content must comply with the following:

• ❌ No explicit, pornographic, or violent content
• ❌ No references to politics, religion, or sensitive events
• ❌ No privacy violations, fraud inducement, or false advertising
• ❌ No covert promotion (e.g., unsolicited links to Discord, X/Twitter, Facebook, or external websites)

Note: External references (e.g., Input template) are acceptable if necessary for context — but aggressive self-promotion or personal contact info is prohibited.

---

​

III. Experience Evaluation Criteria

​

1. Language Consistency

* ( Note: Internal processing may use English for model inference, but final output must strictly adhere to the input language.) ⚠️ Rejection if:

• Unjustified language mixing (e.g., English input → Chinese output)

Note: Language consistency checks are not required for pure images or pure audio/video—only when visible/overlayed text is included.

---

​

2. Functional Availability

​

File Upload Support

⚠️ Rejection if:

• Claims image processing support but fails on JPG, PNG, WebP, JPEG
• The agent claims to support file processing (e.g., Excel, Word, and PDF documents), but the input interface does not allow file uploads.

Note: All upload restrictions (e.g., “1 image only”) must be visible in the input prompt/interface, and enforced in code (e.g. Up to 1 files).

​

Output File Download

• All generated files (images, videos, PDFs, etc.) must be downloadable
• Verification: Click “Download” → confirm successful file save
• File Integrity Check： For text-based files (e.g., PDFs with selectable text, CSV): Ensure content is human-readable with no encoding errors (e.g., symbols), extraneous whitespace, or corrupted characters. For binary files (e.g., JPG, PNG, MP4): Ensure the file is non-empty, not truncated, and renders/opens without errors.

Note: As long as users can access/retrieve the output(e.g. HTMLs), the requirement is satisfied. Note: The image output must be rendered directly at the end of the workflow—not merely provided as a URL.

​

Clickable Output URLs

• All links must be functional and lead to valid, safe destinations
• Must not be broken (e.g., http://error) or redirect to malicious/blacklisted sites

---

​

3. Ease of Use

User input requirements should not be too stringent; minimize the number of input steps and the amount of uploaded materials: ⚠️ Rejection if:

• Entry barrier is excessively high, harming UX
• Requires users to manually edit/Photoshop input images

---

​

4. Minimum Output Quality

Text

• No severe hallucinations (e.g., fabricating non-existent facts)
• No common-sense violations
• Occasional word blocking is acceptable and can be tolerated.

❌ Reject: frequent or patterned censorship URLs

• Links must be real and accessible
• No mass generation of invalid or shortened links

Images

• Resolution ≥ 512×512 px
• Clear, coherent visuals; minimal artifacts/distortions
• Artistic nudity permitted (e.g., classical sculptures); explicit sexual content prohibited
• For outfit-transfer agents: approve if changes are imperceptible (within 10% difference) or visually ambiguous
• ❌ Reject: cropped/incomplete outputs or severely blurred images

Videos

• No glaring physical/logical errors (e.g., walking upside-down)
• Smooth playback; no stuttering or frame tearing

---

​

IV. Common Rejection Cases

• Issue Type: ❌ Vague Information
• Example: Name: “Smart Writing Tool” — no specific functionality described
• Issue Type: ❌ Language Inconsistency
• Example: English input → full Chinese output
• Issue Type: ❌ Missing Functionality
• Example: Claims “PDF parsing supported”, but upload fails with error
• Issue Type: ❌ Low-Quality Output
• Example: Generated image: 256×256, blurry
• Issue Type: ❌ Safety Violation
• Example: Showcase image contains politically sensitive symbols
• Issue Type: ❌ Covert Promotion
• Example: Output includes: “Add me on Telegram: xxx”
• Issue Type: ❌ Insufficient Runtime Stability
• Requirement: Each submitted version must have 5 consecutive successful runs before UTC 23:59 on the day of submission
• Example 1: New version has only 3 successful runs → Rejected
• Example 2: New version has 5 runs but only 4 successful and 1 failed → Rejected (runs must be consecutive)

---

​

Premium Agent Standards

​

I. Agent Classification

Primary classification mainly distinguishes Entertainment from other categories. ⚠️ Entertainment category agents do not require premium review. Classification Criteria:

​

II. Information Review

​

Showcase / Cover

✅ Requirements: Operation/Effect visualization. Use comparison images, videos, GIF animations, playback, etc., to demonstrate usage effects. ✅ Examples: ✔️ Before-after Images ✔️ Playback ✔️ GIF Animation

​

III. Experience Evaluation

​

1. Page Design

✅ Requirements: Excellent page design: All page elements should follow a unified, immersive visual specification. Page loads very quickly, and interactive animations are smooth and lag-free.

​

2. Input Assistance Intelligence

✅ Examples: ✔️ Pre-set parameter tuning with one-click reuse. ✔️ AI model-assisted input. ✔️ Differentiates between professional and beginner input methods.

​

3. Effect Presentation

✅ Requirements: Clear user prompts: ✔️ Correct output must display clear success messages. ✔️ Clear error message guidance. ✔️ Handling solution (auto-retry mechanism). Progress visualization: ✔️ Iframes display real-time progress, or loading animations. ⚠️ Wait time standards vary based on Agent type. ⚠️ N8N automatically scores zero for this item. Complete tasks in an extremely short time: ✔️ Non-image generation tasks output results about in 1 minute. ✔️ Images generated within 5 seconds.

​

4. Functional Scarcity

✅ Requirements: Features new to the platform: ✔️ Examples: MyBill Extractor, TableRay - DOCX Table Autofill Agent. Output exceeds expectations: ✔️ Integrates more upstream/downstream production capabilities. Offers diverse, high-quality options that are easy for users to control and modify. Outputs accurately fulfill user needs even under complex workflows.

To find navigation and other pages in this documentation, fetch the llms.txt file at: https://mulerun.com/docs/llms.txt