{"skill":{"slug":"skylv-prompt-optimization-tool","displayName":"Skylv Prompt Optimization Tool","summary":"Automatically analyzes, A/B tests, and optimizes AI prompts to improve output quality, reduce token usage, and increase consistency.","description":"---\nname: ai-prompt-optimizer\nslug: skylv-ai-prompt-optimizer\nversion: 1.0.2\ndescription: AI prompt auto-optimizer. A/B tests prompts and auto-improves output quality while reducing token costs. Triggers: prompt optimization, prompt engineering, reduce tokens.\nauthor: SKY-lv\nlicense: MIT\ntags: [prompt, optimization, ab-testing, engineering, ai]\nkeywords: prompt, optimization, llm, gpt\ntriggers: ai prompt optimizer\n---\n\n# AI Prompt Optimizer — 提示词自动优化器\n\n## 功能说明\n\n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\"。\n\n## 核心能力\n\n### 1. 提示词分析 (Prompt Analysis)\n\n```yaml\nanalysis_dimensions:\n  - clarity: 清晰度（指令是否明确）\n  - specificity: 具体性（是否有足够细节）\n  - structure: 结构化（是否有清晰格式）\n  - examples: 示例（是否有 few-shot 示例）\n  - constraints: 约束（是否有输出限制）\n  - tone: 语气（是否适合场景）\n\nscoring:\n  - overall_score: 0-100\n  - dimension_scores: 各维度分数\n  - improvement_areas: 需改进的方面\n```\n\n**使用示例：**\n```\n用户：分析这个提示词的质量\nAgent:\n  1. 多维度评分\n  2. 识别问题点\n  3. 提供改进建议\n```\n\n### 2. A/B 测试 (A/B Testing)\n\n```yaml\ntest_setup:\n  - variants: 2-5 个提示词变体\n  - sample_size: 每个变体 100+ 次测试\n  - metrics: 质量评分、Token 消耗、用户满意度\n  - duration: 7-14 天\n\nmetrics_tracked:\n  - output_quality: 输出质量（1-5 分）\n  - token_efficiency: Token 效率\n  - user_satisfaction: 用户满意度\n  - task_completion: 任务完成率\n```\n\n**使用示例：**\n```\n用户：为这个提示词运行 A/B 测试\nAgent:\n  1. 生成 3 个变体\n  2. 分配流量（33% each）\n  3. 收集性能数据\n  4. 选出最优版本\n```\n\n### 3. 自动优化 (Auto Optimization)\n\n```yaml\noptimization_techniques:\n  - prompt_compression: 压缩冗余内容\n  - structure_addition: 添加结构化格式\n  - example_injection: 注入 few-shot 示例\n  - constraint_refinement: 优化约束条件\n  - tone_adjustment: 调整语气风格\n\nexpected_improvements:\n  - token_reduction: 30-60%\n  - quality_improvement: 20-40%\n  - consistency: 提升 50%+\n```\n\n**使用示例：**\n```\n用户：优化这个提示词\nAgent:\n  1. 分析当前版本\n  2. 应用优化技术\n  3. 输出优化版本\n  4. 对比性能数据\n```\n\n### 4. 提示词库 (Prompt Library)\n\n```yaml\ncategories:\n  - writing: 写作类\n  - coding: 编程类\n  - analysis: 分析类\n  - creative: 创意类\n  - business: 商业类\n\nfeatures:\n  - search: 关键词搜索\n  - filter: 按类别/评分筛选\n  - rating: 社区评分\n  - versioning: 版本历史\n```\n\n## 优化框架\n\n### BEFORE → AFTER 对比\n\n**❌ 低效提示词：**\n```\n帮我写一个 Python 函数，要能处理各种情况，\n考虑周全一点，输出要好。\n```\n\n**✅ 优化后：**\n```\n# 角色\nPython 高级开发工程师\n\n# 任务\n编写一个数据验证函数\n\n# 输入\n- data: dict，待验证数据\n- schema: dict，验证规则\n\n# 输出\n- valid: bool，是否通过验证\n- errors: list，错误列表（如有）\n\n# 约束\n- 使用 type hints\n- 添加 docstring\n- 包含单元测试示例\n- 处理边界情况\n\n# 示例\n输入：{\"name\": \"John\", \"age\": 25}\n输出：{\"valid\": True, \"errors\": []}\n```\n\n**效果对比：**\n| 指标 | Before | After | 提升 |\n|------|--------|-------|------|\n| Token 消耗 | 800 | 450 | -44% |\n| 输出质量 | 3.2/5 | 4.6/5 | +44% |\n| 一致性 | 60% | 92% | +53% |\n\n### 优化技巧清单\n\n**1. 角色定义**\n- ❌ \"你是一个助手\"\n- ✅ \"你是拥有 10 年经验的资深 Python 工程师，擅长编写生产级代码\"\n\n**2. 任务明确**\n- ❌ \"帮我处理这个\"\n- ✅ \"分析以下数据，输出 3 个关键洞察，每个洞察包含数据支撑\"\n\n**3. 输出格式**\n- ❌ \"输出结果\"\n- ✅ \"以 JSON 格式输出，包含 keys: summary, insights, recommendations\"\n\n**4. 添加示例**\n- ❌ 无示例\n- ✅ \"输入示例：... 期望输出：...\"\n\n**5. 约束条件**\n- ❌ 无约束\n- ✅ \"不超过 500 字，使用专业术语，避免口语化\"\n\n## 工具函数\n\n### analyze_prompt\n\n```python\ndef analyze_prompt(prompt: str) -> dict:\n    \"\"\"\n    提示词分析\n    \n    Args:\n        prompt: 待分析提示词\n    \n    Returns:\n        {\n            \"overall_score\": 65,\n            \"dimensions\": {\n                \"clarity\": 70,\n                \"specificity\": 55,\n                \"structure\": 60,\n                \"examples\": 40,\n                \"constraints\": 50\n            },\n            \"issues\": [\n                \"缺少角色定义\",\n                \"输出格式不明确\",\n                \"没有示例\"\n            ],\n            \"suggestions\": [\n                \"添加专业角色定义\",\n                \"指定 JSON 输出格式\",\n                \"添加 few-shot 示例\"\n            ]\n        }\n    \"\"\"\n```\n\n### optimize_prompt\n\n```python\ndef optimize_prompt(prompt: str, goal: str = \"quality\") -> dict:\n    \"\"\"\n    提示词优化\n    \n    Args:\n        prompt: 原始提示词\n        goal: 优化目标 (quality|tokens|speed)\n    \n    Returns:\n        {\n            \"original\": {...},\n            \"optimized\": \"优化后的提示词\",\n            \"changes\": [\"添加角色\", \"结构化\", \"添加示例\"],\n            \"expected_improvement\": {\n                \"quality\": \"+35%\",\n                \"tokens\": \"-40%\",\n                \"consistency\": \"+50%\"\n            }\n        }\n    \"\"\"\n```\n\n### run_ab_test\n\n```python\ndef run_ab_test(base_prompt: str, variants: list, iterations: int = 100) -> dict:\n    \"\"\"\n    A/B 测试\n    \n    Args:\n        base_prompt: 基础提示词\n        variants: 变体列表\n        iterations: 测试次数\n    \n    Returns:\n        {\n            \"winner\": \"variant_2\",\n            \"results\": [\n                {\"variant\": \"base\", \"score\": 3.8, \"tokens\": 500},\n                {\"variant\": \"v1\", \"score\": 4.1, \"tokens\": 450},\n                {\"variant\": \"v2\", \"score\": 4.6, \"tokens\": 420}\n            ],\n            \"statistical_significance\": 0.95\n        }\n    \"\"\"\n```\n\n### generate_variants\n\n```python\ndef generate_variants(prompt: str, count: int = 5) -> list:\n    \"\"\"\n    生成提示词变体\n    \n    Args:\n        prompt: 原始提示词\n        count: 生成数量\n    \n    Returns:\n        [\n            {\"id\": \"v1\", \"prompt\": \"...\", \"changes\": [\"添加角色\"]},\n            {\"id\": \"v2\", \"prompt\": \"...\", \"changes\": [\"结构化\"]},\n            ...\n        ]\n    \"\"\"\n```\n\n## 提示词模板库\n\n### 写作类\n\n```yaml\ntemplate: blog_post\nprompt: |\n  # 角色\n  资深内容创作者，10 年科技博客经验\n\n  # 任务\n  撰写一篇关于{主题}的博客文章\n\n  # 要求\n  - 字数：2000-2500 字\n  - 结构：引言 + 3-5 个主体段落 + 结论\n  - 语气：专业但易懂\n  - 包含：实际案例、数据支撑、行动建议\n\n  # 输出格式\n  Markdown，包含 H2/H3标题、列表、引用块\n```\n\n### 编程类\n\n```yaml\ntemplate: code_review\nprompt: |\n  # 角色\n  资深代码审查工程师，精通{语言}\n\n  # 任务\n  审查以下代码，输出审查报告\n\n  # 审查维度\n  1. 代码质量（命名、结构、注释）\n  2. 安全性（OWASP Top 10）\n  3. 性能（时间/空间复杂度）\n  4. 可维护性（测试、文档）\n\n  # 输出格式\n  JSON:\n  {\n    \"issues\": [{\"severity\": \"high|medium|low\", \"description\": \"...\", \"fix\": \"...\"}],\n    \"score\": 0-100,\n    \"summary\": \"...\"\n  }\n```\n\n### 分析类\n\n```yaml\ntemplate: data_analysis\nprompt: |\n  # 角色\n  数据科学家，擅长商业洞察\n\n  # 任务\n  分析以下数据集，输出商业洞察\n\n  # 分析框架\n  1. 描述性统计（均值、中位数、分布）\n  2. 趋势分析（同比、环比）\n  3. 异常检测（离群值、异常模式）\n  4. 商业建议（可行动洞察）\n\n  # 输出格式\n  Markdown 报告，包含图表描述、关键数字、行动建议\n```\n\n## 相关文件\n\n- [Prompt Engineering Guide](https://www.promptingguide.ai)\n- [OpenAI Best Practices](https://platform.openai.com/docs/guides/prompt-engineering)\n- [Anthropic Prompt Design](https://docs.anthropic.com/claude/docs/prompt-design)\n\n## 触发词\n\n- 自动：检测 prompt、optimize、improve、A/B testing 相关关键词\n- 手动：/prompt-optimizer, /optimize-prompt, /ab-test\n- 短语：优化提示词、改进 prompt、A/B 测试\n\n## Usage\n\n1. Install the skill\n2. Configure as needed\n3. Run with OpenClaw\n","tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":321,"installsAllTime":1,"installsCurrent":1,"stars":0,"versions":1},"createdAt":1777783752926,"updatedAt":1778492833926},"latestVersion":{"version":"1.0.0","createdAt":1777783752926,"changelog":"- Initial release of AI Prompt Optimizer.\n- Automatically analyzes and optimizes AI prompts to improve output quality and reduce token usage using A/B testing and performance data.\n- Features include multi-dimensional prompt analysis, automated A/B prompt testing, and intelligent optimization suggestions.\n- Provides a structured prompt template library for writing, coding, and data analysis use cases.\n- Includes utility functions for analyzing, optimizing, generating variants, and testing prompts.","license":"MIT-0"},"metadata":null,"owner":{"handle":"sky-lv","userId":"s17fgkeb63szvtadtmm753m0gd84e4vz","displayName":"SKY-lv","image":"https://avatars.githubusercontent.com/u/259750852?v=4"},"moderation":null}