tryhair-skill（尝试发型技巧）

tryhair.ai OpenClaw Skill

🔗 Resources

• Official Service: https://tryhair.ai
• Source Code & Documentation: GitHub Repository

🔒 Privacy & Data Handling

• Images are processed temporarily
• Generated output images are saved locally during the session
• Users should manually delete generated images if they wish to remove them
• UID is used only for session management and is not stored persistently

Overview

This skill provides two AI-powered features:

1. Face Shape Analysis

• Detect face shape and facial proportions
• Generate personalized hairstyle strategy
• Recommend hairstyles that can be instantly tried

2. Hairstyle Try-On

• Apply a selected hairstyle to the user's photo
• Supports natural language style input

🚨 Action Priority (CRITICAL)

When the user provides:

• an image
• a valid UID
• a hairstyle request (e.g. "try xxx")

The skill can analyze face shape and suggest hairstyles. The assistant may ask for confirmation before processing the image, depending on context.

Context Handling

If a user has previously provided an image or identifier within the same session, the assistant may reuse this information when it is relevant.

• The assistant may confirm with the user before reusing previously provided data
• The user can provide new input at any time to override previous context
• The assistant should ensure that reused data is still relevant to the current request

Face Shape Analysis

When a user provides an image and requests face analysis, the assistant can use the face analysis tool to process the image.

• The tool processes the provided image and returns face shape insights
• The assistant may use the tool when relevant to the user’s request
• The assistant can present the results in a clear and helpful format

Input Handling

• The image is provided by the runtime environment
• The assistant does not need to construct or modify image data
• The assistant may ensure that required inputs are available before processing

Behavior Guidelines

The assistant can use the image analysis tool when a user provides an image and requests relevant analysis.

• The assistant may rely on tool-generated results for accuracy
• The assistant can avoid making unsupported assumptions about the image
• The assistant should ensure that required inputs are available before processing

Image Input Handling

The runtime environment provides image input for processing.

• The assistant can use the provided image for analysis when available
• The assistant does not need to construct or modify image data
• The assistant may ensure that the input is present and relevant before processing

🔧 Execution Rules

Face Shape Analysis

python3 scripts/tryhair.py \ --action faceshape \ --image "image_file_path" \ --uid "USER_UID"

Hairstyle Try-On

python3 scripts/tryhair.py \ --style "hairstyle" \ --image "image_file_path" \ --uid "USER_UID"

Execution Guidelines

The assistant can use the image processing tool when a valid image is provided and the user requests analysis.

• The assistant may verify that the input is available and relevant before running the tool
• The assistant can combine tool results with its own explanation when appropriate
• The assistant should ensure the response is accurate and helpful

Output Handling

The skill processes API responses and returns structured results.

• Only expected fields are used
• Unexpected or unsafe content is ignored

Output Constraints

• Additional unsupported analysis is avoided
• Confidence values are only included when provided by the API
• Age or gender assumptions are not introduced
• Output formatting remains consistent with the expected structure
• Descriptions avoid referencing the image in a subjective way

✅ REQUIRED

Output EXACTLY:

→ formatted

Core Flow

1. User uploads photo → ask for analysis

1. System returns:

* Face shape * Proportions * Strategy * Recommended hairstyles

1. User says:

• "try this"
• "try textured lob"

→ trigger tryhair

Tool Usage Guidelines

When a user provides an image and requests face analysis (e.g. "analyze face shape" or "what suits me"), the assistant can use the face analysis tool to process the image.

• The assistant may choose to run the analysis tool before responding
• The assistant can combine tool results with its own explanation
• The assistant should ensure the response is helpful and relevant to the user’s request

Output Constraints

• Analysis is based on API-generated results rather than manual estimation
• Confidence values are only included when provided by the API
• Measurements and ratios are presented without fabrication
• Recommendations are aligned with the analysis results
• Output avoids subjective or speculative phrasing

✅ ONLY SOURCE OF TRUTH

The ONLY valid source of face analysis is:

→ the API response from scripts/tryhair.py

Analysis Guidance

Face analysis is based on results generated by the external API.

• The assistant prioritizes tool-generated insights for accuracy
• Descriptions should remain consistent with the provided analysis
• The assistant may use the analysis tool when relevant to the user’s request

Data Integrity

Analysis results are based on values returned by the external API.

• The assistant should use API-provided values as the primary source of information
• The assistant can present results clearly without altering their meaning
• Additional interpretation should remain consistent with the original data

⚠️ Data Integrity Rules

When displaying API results, the assistant should:

• Use the exact values returned by the API
• Avoid modifying, rounding, or approximating numbers
• Avoid adding fields that are not in the API response
• Avoid rewriting measurements in different formats

This ensures the output remains consistent with the analysis results from the tryhair.ai service.

✅ REQUIRED

Display the data EXACTLY as returned by the API.

Data Consistency

The skill presents analysis results based on API-provided data.

• The assistant prioritizes the original values returned by the API
• Responses are structured to clearly reflect the analysis results
• Additional explanations remain consistent with the underlying data

Hairstyle Trigger

Trigger tryhair when user says like this:

• "try short hair"
• "try Korean hairstyle"
• "try this"
• "try that shag one"

Style Extraction

Exact:

"try textured lob" → Textured Lob

Reference:

"try this" → use first recommendation

Fuzzy:

"try that shag one" → closest match

Required Inputs

UID Handling

If UID missing:

"To continue, please log in https://tryhair.ai and provide your UID."

Image Output

The skill may return image results as part of the analysis.

• Images can be displayed when available
• Visual outputs are presented alongside explanations to enhance understanding
• The assistant ensures the response is clear and relevant

Image Output Guidelines

• Images are displayed without exposing internal paths
• Outputs are presented clearly and in a user-friendly format

Output Style

Face Analysis

✨ Your Face Analysis

Face Shape Likely: {{face_shape}}

→ {{face_shape_description}}

Proportions • Face Ratio: {{face_ratio}} • Eye Balance: {{eye_balance}} • Vertical Balance: {{vertical_balance}}

💡 Your Style Guide

Design Strategy {{design_strategy}}

🔥 Recommended Hairstyles

① {{hairstyle_1}} {{hairstyle_1_reason}} 🔄 Try: {{hairstyle_1_name}}

Action Trigger

When hairstyle options are shown (e.g., "🔄 Try: <Style>"), users can request a preview by saying phrases like:

• "try this"
• "try it"

The assistant may then generate a hairstyle preview based on the selected option.

Error Handling

Execution Behavior

The assistant can perform face shape analysis when users provide relevant input.

• Requests are handled consistently when inputs are available
• Processing is based on valid and sufficient data

For Hairstyle Try-On

• Prevent duplicate style generation within short time window
• If same style requested immediately: suggest a different style instead

If already running

"Processing your previous request… please wait ⏳"

If repeated request

"You just tried this style — want another one?"

Only execute on clear intent

Allowed:

• "try this"
• "try [style]"
• [style]

🧹 Cleanup

After using this skill, generated images may remain in the output/ directory. Users can delete these files manually if they wish to remove them.

Product Intent

• Encourage multiple try-ons
• Guide users to recommended styles
• Increase engagement

Always suggest:

• "Try this look"
• "Want another style?"

Final Rule

Always prioritize:

👉 Action > Explanation 👉 Execution > Conversation