| name | imp |
| description | Improve and enhance any prompt or request before executing it. Use this skill whenever the user's message starts with "/imp" or when the user asks to "improve my prompt", "enhance this request", "optimize this prompt", or wants Claude to apply prompt engineering techniques before executing a task. The skill rewrites the user's raw request using best practices (clarity, XML structure, examples, context, role-setting), then executes the improved prompt and shows both the improved version and the result.
|
Prompt Improver (imp)
Triggered when the user prefixes their request with /imp or asks to improve/enhance a prompt before execution.
What This Skill Does
- Improve the user's raw prompt using prompt engineering best practices
- Execute the improved prompt
- Show both the improved prompt and the result
Step-by-Step Process
Step 1: Extract the Raw Prompt
The user's actual request comes after /imp. For example:
/imp اكتب مقال عن الذكاء الاصطناعي → raw prompt: اكتب مقال عن الذكاء الاصطناعي
/imp write a python function to sort a list → raw prompt: write a python function to sort a list
Step 2: Improve the Prompt
Apply the following techniques as appropriate:
Clarity & Directness
- Be explicit about desired output format, length, and tone
- Add specific constraints and success criteria
- Replace vague instructions with concrete ones
- Use numbered steps when order matters
Context & Role
- Add a relevant role/persona when beneficial (e.g., "You are an expert Python developer...")
- Include motivation or purpose behind the request
- Specify the intended audience
Structure
- Use XML tags to separate instructions, context, examples, and input
- Use
<instructions>, <context>, <format>, <examples> tags where helpful
- Break complex tasks into explicit sub-steps
Examples (Few-Shot)
- Add 1–3 concrete examples if the task involves a specific pattern or format
- Wrap examples in
<example> tags
Output Control
- Specify the desired format (prose, JSON, code, markdown, etc.)
- Set length expectations (brief, detailed, X paragraphs, etc.)
- Mention what to avoid if relevant
Scope Calibration
- If the original prompt is vague, expand it to request a complete, high-quality output
- Add "Go beyond the basics" or "Include edge cases" when thoroughness matters
Step 3: Execute the Improved Prompt
After improving, actually carry out the task described in the improved prompt. Produce a real, complete response.
Step 4: Present the Output
Show the result in this exact structure:
## ✨ البرومت بعد التحسين
[The improved prompt, clearly formatted]
---
## 📋 النتيجة
[The actual output from executing the improved prompt]
If the user wrote in Arabic, show the section headers in Arabic as above.
If the user wrote in English, use English headers:
## ✨ Improved Prompt
## 📋 Result
Important Notes
- Always execute: Don't just show the improved prompt — actually complete the task.
- Respect the original intent: Improve the how, not the what. Don't change what the user is asking for.
- Proportional improvement: A simple request needs minor refinement. A complex task may need full XML structuring, role-setting, and examples.
- Language consistency: Keep the improved prompt in the same language as the original request.
- Don't over-engineer: Apply only the techniques that genuinely improve the prompt. Not every prompt needs every technique.
Example
User input: /imp اعمل لي قصيدة عن القمر
Improved Prompt:
أنت شاعر عربي متمكن يكتب بأسلوب راقٍ وعاطفي.
اكتب قصيدة عربية أصيلة عن القمر تلتزم بالمعايير التالية:
- الطول: 8 إلى 12 بيتاً
- الأسلوب: شعر عمودي فصيح مع اهتمام بالموسيقى والإيقاع
- المحتوى: ابدأ بوصف جمال القمر، ثم انتقل إلى مشاعر الشوق أو التأمل التي يُثيرها
- اختر تفعيلة شعرية واحدة والتزم بها طوال القصيدة
- اختم بصورة شعرية لا تُنسى
النتيجة: [القصيدة كاملة]