prompt-assistant

Files (15)

• LICENSE.txt
• README.md
• SKILL.md
• anti-patterns.md · reference/
• best-practices.md · reference/
• common-patterns.md · reference/
• tools-specific-prompts.md · reference/
• complex-decomposition-template.md · templates/
• data-parsing-template.md · templates/
• financial-rule-analysis-prompt.md · templates/
• financial-rule-output-example.md · templates/
• financial-rule-output-template.md · templates/
• markdown-output-template.md · templates/
• prompt-optimization-guide.md · templates/
• role-based-system-prompt-template.md · templates/

SKILL.md (first lines)

⚠️ **PLEASE READ THIS ENTIRE FILE BEFORE GENERATING PROMPTS** ⚠️

# Prompt Assistant Skill

Help users generate, optimize, and adapt high-quality AI prompts for diverse use cases. This skill provides **specialized templates** for different scenarios, ensuring prompts are precise, structured, and effective.

## Overview

This skill helps with:
- 🎯 **Prompt generation** - Create effective prompts from scratch
- 🔄 **Prompt optimization** - Refine existing prompts for better results
- 📋 **Template selection** - Match the right template to the use case
- 📊 **Structured outputs** - Parse data with consistent formatting
- 🔪 **Complex decomposition** - Break down multi-step tasks
- 📝 **Markdown formatting** - Generate well-structured markdown with sections, content, and diagrams

## When to Use This Skill

**Use for:**
✅ Generating prompts for data parsing and extraction
✅ Creating multi-step instruction prompts
✅ Building prompts that require markdown structured output
✅ Optimizing prompts for better AI responses
✅ Adapting prompts for different tools/models
✅ Creating system prompts and role-based prompts

**Don't use for:**
❌ Direct task execution (use domain-specific skills instead)
❌ General conversation (not needed for simple questions)
❌ Content that violates policies

---

## Core Concept: Scenario-Based Templates

This skill uses **4 primary template categories** based on different use cases:

### 1️⃣ **Data Parsing Template**
Use when the user needs **structured, consistent output** from unstructured data.

**Characteristics:**
- Requires precise data extraction
- Output needs schema/format validation
- Often involves JSON, CSV, or table formats
- Example use cases: Invoice parsing, entity extraction, log analysis

**Key prompt elements:**
```markdown
## Role & Context
Define the parser's expertise and context

## Input Specification
Describe the format of incoming data

## Output Schema
Specify exact output structure with examples

## Parsing Rules
- Rule 1: ...
- Rule 2: ...
- Rule 3: ...

## Error Handling
Instructions for ambiguous or invalid data

## Examples
- Example 1: [Input] → [Expected Output]
- Example 2: [Input] → [Expected Output]
```

### 2️⃣ **Complex Task Decomposition Template**
Use when the user needs **step-by-step instructions** for multi-part problems.

**Characteristics:**
- Large task with multiple phases
- Sequential dependencies between steps
- Requires intermediate validation
- Example use cases: Content creation, analysis, research planning

**Key prompt elements:**
```markdown
## Objective
Clear, measurable end goal

## Phase Overview
High-level breakdown of major phases

## Detailed Steps
Phase 1: [Step 1a, 1b, 1c...]
Phase 2: [Step 2a, 2b, 2c...]
Phase 3: [Step 3a, 3b, 3c...]

## Success Criteria
How to validate each phase

## Key Considerations
Important context and constraints

## Fallback Options
Alternative approaches if something fails
```

### 3️⃣ **Markdown Structured Output Template**
Use when the user needs **formatted markdown output** with specific sections, hierarchy, and visual elements.

**Characteristics:**
- Output must be well-structured markdown
- Different section types (narrative, code, diagrams)
- Visual hierarchy and readability important
- Example use cases: Documentation, reports, educational content

**Key prompt elements:**
```markdown
## Output Structure
Section 1: [Description]
Section 2: [Description]
Section 3: [Description]

## Formatting Rules

From zhaoxuanZzz/skills, MIT.

More from zhaoxuanZzz/skills