📦 WheelSpotter

WheelSpotter (v1.0)

🎯 WheelSpotter — Your wheel-spotting scout. Spots reusable solutions before you build from scratch.

Core principle: Solutions must be directly integrable—not flashy but unusable toys.

When to Use ✅ Trigger Scenarios

Load this 技能 when the user expresses these intents:

Pattern Example Looking for existing solutions "Is there an existing PDF parsing 库?" Avoiding duplicate work "I don't want to reinvent the wheel..." Tech stack consultation "What's a good Python data 可视化 库?" Quick integration needs "I need an OCR API I can use right away" Pre-implementation re搜索 "Implementing JWT auth—any existing solutions?" Wheel spotting "Spot any wheels for image processing?"

Keyword matches: is there, existing, wheel, 库, 框架, API, 工具, solution, spot

❌ Do NOT Trigger Scenario Reason Suggestion User wants to build themselves "I want to write my own..." Assist with coding directly Highly customized requirements "I need something that does X, Y, Z all at once..." Suggest breaking down and 搜索ing separately Learning purposes "I want to learn how to implement..." Provide tutorials instead Tech stack already decided "I'm using React to build..." Move to development 图形界面dance De签名 Principles Principle Description Implementation Problem-Oriented Precisely solve "finding integrable wheels" Sources classified by 输出 form, exclude chat机器人s Closed-Loop Delivery Clear "usable/unusable" conclusion with action 结果s include pip 安装 commands or self-build recommendation High Adaptability Dynamic strategy based on complexity and intent Complexity grading + intent-adaptive source selection 进度ive Improvement 系统 获取s smarter with each use Feedback loops, 结果 caching, vector memory Transferable Leverage Core capabilities reusable elsewhere Funnel engine, cost 监控 as independent 模块s Cost Red Line 搜索 cost must be lower than self-build cost Bud获取 caps, tiered abandonment, early termination Prerequisites pip 安装 -r requirements.txt

环境:

Python 3.8+ Internet 访问 for API calls GitHub 令牌 (optional, increases API limit to 5000 req/hour) 输入/输出 Specification 输入 格式化 # Method 1: Natural language (解析d by 代理) user_输入 = "I need a Python 库 to process Excel files"

# Method 2: Structured 输入 (optional) { "requirement": "process Excel files", "tech_stack": ["Python"], "intent": "库", "constrAInts": { "license": "MIT", "min_stars": 100, "last_更新d": "12m" } }

输出 格式化 { "状态": "found", "recommendations": [ { "name": "openpyxl", "source": "pypi", "url": "https://openpyxl.readthedocs.io/", "match_score": 0.92, "integration_score": 0.95, "action": "pip 安装 openpyxl", "license": "MIT", "stars": 1200, "last_更新d": "2 months ago", "警告s": [], "advice": "Recommended, mature and stable" } ], "fallback": null, "cost": { "令牌s_used": 420, "time_seconds": 3.2, "estimated_time_saved": "~4 hours" } }

状态 values:

found: Suitable solutions found not_found: Recommend self-build needs_clarification: Requirement unclear, need follow-up error: 搜索 fAIled, return error 信息 Core 工作流 User 输入 ↓ [M0] Complexity Grading (~30 令牌s) ↓ [M1] Intent Classification (~60 令牌s) ↓ [Optional] Clarification (1-2 rounds if needed) ↓ [M2] 提取 Keywords + Tech Entities (~150 令牌s) ↓ [搜索] Activate 平台s by intent, parallel API calls ↓ [Hard 过滤器] Deprecated/activity/form matching ↓ [LLM Refinement] Multi-dimensional eval for ≤5 candidates (~300 令牌s) ↓ 输出 recommendations + action commands + cost 报告

Implementation DetAIls Step 1: Complexity Grading (M0)

Prompt Template:

You are a development complexity assessment expert. Evaluate the requirement:

• L1: Simple function/工具, solvable with dozens of lines
• L2: Medium 模块, requires interface de签名
• L3: Complex 系统, involves multiple 组件s

Requirement: {requirement} 输出 JSON only: { "complexity": "L2", "reason": "..." }

Impact on 搜索 Strategy:

Complexity 令牌 Cap Time Cap Sources Star Threshold L1 Simple 300 8s 2-3 ≥10 L2 Medium 600 12s 3-5 ≥50 L3 Complex 800 15s Full ≥100 Step 2: Intent Classification (M1)

Prompt Template:

Analyze the requirement, determine desired 输出 form (multiple allowed):

• 库: 库/框架 integrable into code
• 服务: Callable external API/服务
• 工具: Standalone executable 工具/命令行工具
• reference: Code template/example/architecture reference
• 助手: Conversational 助手 (usually not a wheel, use cautiously)

Requirement: {requirement} 输出 JSON only: { "intent": [...], "reason": "..." }

导入ant: If intent only contAIns 助手, return 图形界面dance without triggering 搜索.

Step 3: Platfor