AgentSkillsCN

improve-prompt

将原始提示词转化为优化后的生产就绪提示词的提示工程架构师。主动启用以下场景：(1) 改进/优化提示词；(2) 针对特定模型优化提示词；(3) 构建提示词链或提示词序列；(4) 分析提示词质量；(5) 根据不同场景转换提示词。 触发指令包括：“改进这个提示词”“优化提示词”“让这个提示词更好”“提示工程”“修复我的提示词”。

SKILL.md
--- frontmatter
name: improve-prompt
description: >
  Prompt engineering architect that transforms raw prompts into optimized,
  production-ready prompts. PROACTIVELY activate for: (1) Improve/refine prompts,
  (2) Optimize prompts for specific models, (3) Create prompt chains/sequences,
  (4) Analyze prompt quality, (5) Transform prompts for different contexts.
  
  Triggers: "improve this prompt", "refine prompt", "optimize prompt",
  "make this prompt better", "prompt engineering", "fix my prompt"

Improve Prompt

Transform prompts into optimized, production-ready versions.

Purpose

Analyze and enhance prompts by:

  • •Identifying weaknesses and ambiguities
  • •Applying prompt engineering best practices
  • •Optimizing for specific models (Claude, Gemini, GPT)
  • •Structuring for consistent outputs
  • •Adding appropriate constraints and examples

When to Use

Ideal for:

  • •Prompts that get inconsistent results
  • •Prompts that need to work across models
  • •Complex prompts requiring structure
  • •Prompts for production use

Avoid when:

  • •Simple, one-off questions
  • •Conversational queries

Workflow

Step 1: Analyze Current Prompt

Evaluate the prompt for:

  • •Clarity: Is the task unambiguous?
  • •Completeness: Is all necessary context provided?
  • •Structure: Is it well-organized?
  • •Constraints: Are boundaries defined?
  • •Examples: Are examples provided if needed?
  • •Output format: Is expected output specified?

Step 2: Identify Improvements

Common issues to address:

  • •Vague instructions → Make specific
  • •Missing context → Add relevant background
  • •No output format → Specify structure
  • •No examples → Add few-shot examples
  • •No constraints → Add guardrails
  • •Too long → Consolidate and prioritize

Step 3: Apply Best Practices

For Claude:

  • •Use XML tags for structure
  • •Leverage extended thinking for complex tasks
  • •Request step-by-step reasoning
  • •Use positive framing ("do X" not "don't do Y")

For all models:

  • •Put important instructions at start and end
  • •Use numbered steps for sequences
  • •Provide concrete examples
  • •Specify output format explicitly
  • •Include edge case handling

Step 4: Generate Improved Prompt

Output the optimized prompt with:

  • •Clear structure
  • •Explicit instructions
  • •Defined output format
  • •Examples if beneficial

Output Format

markdown
## Prompt Analysis

**Original prompt issues:**
1. [Issue 1]
2. [Issue 2]

**Improvements applied:**
1. [Improvement 1]
2. [Improvement 2]

---

## Improved Prompt

[The optimized prompt, ready to copy]

---

## Usage Notes

- **Best for:** [model/use case]
- **Expected output:** [description]
- **Variations:** [any suggested variations]

Quality Gates

  • • All ambiguities resolved
  • • Output format specified
  • • Appropriate length (not bloated)
  • • Tested mentally for edge cases
  • • Model-appropriate techniques used

Examples

Example: Vague prompt improvement

Before: Summarize this document

After: Summarize the following document in 3-5 bullet points. Focus on:

Key findings or conclusions Important data points Recommended actions

Format each bullet as: [Topic]: [1-2 sentence summary] <document> [Document content here] </document>

code