Prompt Optimization Tool

v1.0.0

Automatically analyzes, A/B tests, and optimizes AI prompts to improve output quality while reducing token usage and ensuring consistency.

0· 51·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for sky-lv/prompt-optimization-tool.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Prompt Optimization Tool" (sky-lv/prompt-optimization-tool) from ClawHub.
Skill page: https://clawhub.ai/sky-lv/prompt-optimization-tool
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install prompt-optimization-tool

ClawHub CLI

Package manager switcher

npx clawhub@latest install prompt-optimization-tool
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
The name/description (prompt analysis, A/B testing, optimization, prompt library) align with the SKILL.md content. The skill is purely instructional and does not request unrelated binaries, credentials, or config paths.
Instruction Scope
SKILL.md contains detailed analysis and function signatures (analyze_prompt, optimize_prompt, run_ab_test, generate_variants) and prescribes A/B testing workflows. It does not instruct the agent to read system files, environment variables, or external endpoints, but it assumes the ability to run large-scale A/B tests and gather metrics (user satisfaction, token counts, statistical significance) without describing how to collect telemetry or route traffic. That is an operational gap (not a direct security concern) — the agent will be limited to producing suggestions unless integrated with telemetry/traffic sources.
Install Mechanism
No install spec and no code files — instruction-only skill. Lowest install risk; nothing will be written to disk by the skill itself.
Credentials
No required environment variables, credentials, or config paths are declared. The skill does not request secrets or access beyond standard agent capabilities.
Persistence & Privilege
Default privileges (not always:true, agent-invocable) are used. The skill does not request persistent system privileges or modification of other skills' configs.
Assessment
This skill is an instruction-only prompt-engineering guide: it will produce analyses, optimized prompts, and suggested A/B test setups, but it does not itself run traffic, collect telemetry, or require any credentials. If you expect automated, large-scale A/B testing, you will need to wire the agent to your telemetry/serving infrastructure (and consider privacy and consent for user data). There are no obvious permission or credential red flags, but verify any integrations you build to collect metrics (endpoints, logs, analytics) before connecting them, and be skeptical of the optimistic improvement percentages — treat them as estimates, not guarantees.

Like a lobster shell, security has layers — review code before you run it.

latestvk976c6n5dxyjgcsk22netaant185m9hd
51downloads
0stars
1versions
Updated 18h ago
v1.0.0
MIT-0

AI Prompt Optimizer — 提示词自动优化器

功能说明

基于 A/B 测试和性能数据,自动优化 AI 提示词,提升输出质量并降低 Token 消耗。让提示词工程从"艺术"变成"科学"。

核心能力

1. 提示词分析 (Prompt Analysis)

analysis_dimensions:
  - clarity: 清晰度(指令是否明确)
  - specificity: 具体性(是否有足够细节)
  - structure: 结构化(是否有清晰格式)
  - examples: 示例(是否有 few-shot 示例)
  - constraints: 约束(是否有输出限制)
  - tone: 语气(是否适合场景)

scoring:
  - overall_score: 0-100
  - dimension_scores: 各维度分数
  - improvement_areas: 需改进的方面

使用示例:

用户:分析这个提示词的质量
Agent:
  1. 多维度评分
  2. 识别问题点
  3. 提供改进建议

2. A/B 测试 (A/B Testing)

test_setup:
  - variants: 2-5 个提示词变体
  - sample_size: 每个变体 100+ 次测试
  - metrics: 质量评分、Token 消耗、用户满意度
  - duration: 7-14 天

metrics_tracked:
  - output_quality: 输出质量(1-5 分)
  - token_efficiency: Token 效率
  - user_satisfaction: 用户满意度
  - task_completion: 任务完成率

使用示例:

用户:为这个提示词运行 A/B 测试
Agent:
  1. 生成 3 个变体
  2. 分配流量(33% each)
  3. 收集性能数据
  4. 选出最优版本

3. 自动优化 (Auto Optimization)

optimization_techniques:
  - prompt_compression: 压缩冗余内容
  - structure_addition: 添加结构化格式
  - example_injection: 注入 few-shot 示例
  - constraint_refinement: 优化约束条件
  - tone_adjustment: 调整语气风格

expected_improvements:
  - token_reduction: 30-60%
  - quality_improvement: 20-40%
  - consistency: 提升 50%+

使用示例:

用户:优化这个提示词
Agent:
  1. 分析当前版本
  2. 应用优化技术
  3. 输出优化版本
  4. 对比性能数据

4. 提示词库 (Prompt Library)

categories:
  - writing: 写作类
  - coding: 编程类
  - analysis: 分析类
  - creative: 创意类
  - business: 商业类

features:
  - search: 关键词搜索
  - filter: 按类别/评分筛选
  - rating: 社区评分
  - versioning: 版本历史

优化框架

BEFORE → AFTER 对比

❌ 低效提示词:

帮我写一个 Python 函数,要能处理各种情况,
考虑周全一点,输出要好。

✅ 优化后:

# 角色
Python 高级开发工程师

# 任务
编写一个数据验证函数

# 输入
- data: dict,待验证数据
- schema: dict,验证规则

# 输出
- valid: bool,是否通过验证
- errors: list,错误列表(如有)

# 约束
- 使用 type hints
- 添加 docstring
- 包含单元测试示例
- 处理边界情况

# 示例
输入:{"name": "John", "age": 25}
输出:{"valid": True, "errors": []}

效果对比:

指标BeforeAfter提升
Token 消耗800450-44%
输出质量3.2/54.6/5+44%
一致性60%92%+53%

优化技巧清单

1. 角色定义

  • ❌ "你是一个助手"
  • ✅ "你是拥有 10 年经验的资深 Python 工程师,擅长编写生产级代码"

2. 任务明确

  • ❌ "帮我处理这个"
  • ✅ "分析以下数据,输出 3 个关键洞察,每个洞察包含数据支撑"

3. 输出格式

  • ❌ "输出结果"
  • ✅ "以 JSON 格式输出,包含 keys: summary, insights, recommendations"

4. 添加示例

  • ❌ 无示例
  • ✅ "输入示例:... 期望输出:..."

5. 约束条件

  • ❌ 无约束
  • ✅ "不超过 500 字,使用专业术语,避免口语化"

工具函数

analyze_prompt

def analyze_prompt(prompt: str) -> dict:
    """
    提示词分析
    
    Args:
        prompt: 待分析提示词
    
    Returns:
        {
            "overall_score": 65,
            "dimensions": {
                "clarity": 70,
                "specificity": 55,
                "structure": 60,
                "examples": 40,
                "constraints": 50
            },
            "issues": [
                "缺少角色定义",
                "输出格式不明确",
                "没有示例"
            ],
            "suggestions": [
                "添加专业角色定义",
                "指定 JSON 输出格式",
                "添加 few-shot 示例"
            ]
        }
    """

optimize_prompt

def optimize_prompt(prompt: str, goal: str = "quality") -> dict:
    """
    提示词优化
    
    Args:
        prompt: 原始提示词
        goal: 优化目标 (quality|tokens|speed)
    
    Returns:
        {
            "original": {...},
            "optimized": "优化后的提示词",
            "changes": ["添加角色", "结构化", "添加示例"],
            "expected_improvement": {
                "quality": "+35%",
                "tokens": "-40%",
                "consistency": "+50%"
            }
        }
    """

run_ab_test

def run_ab_test(base_prompt: str, variants: list, iterations: int = 100) -> dict:
    """
    A/B 测试
    
    Args:
        base_prompt: 基础提示词
        variants: 变体列表
        iterations: 测试次数
    
    Returns:
        {
            "winner": "variant_2",
            "results": [
                {"variant": "base", "score": 3.8, "tokens": 500},
                {"variant": "v1", "score": 4.1, "tokens": 450},
                {"variant": "v2", "score": 4.6, "tokens": 420}
            ],
            "statistical_significance": 0.95
        }
    """

generate_variants

def generate_variants(prompt: str, count: int = 5) -> list:
    """
    生成提示词变体
    
    Args:
        prompt: 原始提示词
        count: 生成数量
    
    Returns:
        [
            {"id": "v1", "prompt": "...", "changes": ["添加角色"]},
            {"id": "v2", "prompt": "...", "changes": ["结构化"]},
            ...
        ]
    """

提示词模板库

写作类

template: blog_post
prompt: |
  # 角色
  资深内容创作者,10 年科技博客经验

  # 任务
  撰写一篇关于{主题}的博客文章

  # 要求
  - 字数:2000-2500 字
  - 结构:引言 + 3-5 个主体段落 + 结论
  - 语气:专业但易懂
  - 包含:实际案例、数据支撑、行动建议

  # 输出格式
  Markdown,包含 H2/H3标题、列表、引用块

编程类

template: code_review
prompt: |
  # 角色
  资深代码审查工程师,精通{语言}

  # 任务
  审查以下代码,输出审查报告

  # 审查维度
  1. 代码质量(命名、结构、注释)
  2. 安全性(OWASP Top 10)
  3. 性能(时间/空间复杂度)
  4. 可维护性(测试、文档)

  # 输出格式
  JSON:
  {
    "issues": [{"severity": "high|medium|low", "description": "...", "fix": "..."}],
    "score": 0-100,
    "summary": "..."
  }

分析类

template: data_analysis
prompt: |
  # 角色
  数据科学家,擅长商业洞察

  # 任务
  分析以下数据集,输出商业洞察

  # 分析框架
  1. 描述性统计(均值、中位数、分布)
  2. 趋势分析(同比、环比)
  3. 异常检测(离群值、异常模式)
  4. 商业建议(可行动洞察)

  # 输出格式
  Markdown 报告,包含图表描述、关键数字、行动建议

相关文件

触发词

  • 自动:检测 prompt、optimize、improve、A/B testing 相关关键词
  • 手动:/prompt-optimizer, /optimize-prompt, /ab-test
  • 短语:优化提示词、改进 prompt、A/B 测试

Usage

  1. Install the skill
  2. Configure as needed
  3. Run with OpenClaw

Comments

Loading comments...