From 096621eee7b418b8e390f8477c428881e6257ba5 Mon Sep 17 00:00:00 2001 From: catlog22 Date: Thu, 6 Nov 2025 16:56:34 +0800 Subject: [PATCH] =?UTF-8?q?docs:=20=E5=BC=BA=E8=B0=83=20SKILL=20=E6=99=BA?= =?UTF-8?q?=E8=83=BD=E6=95=B4=E5=90=88=E5=8E=9F=E5=88=99=EF=BC=8C=E9=81=BF?= =?UTF-8?q?=E5=85=8D=E6=A8=A1=E6=9D=BF=E7=9B=B4=E6=8E=A5=E5=A4=8D=E5=88=B6?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 核心修改： 1. 新增"Core Principle: Intelligent Integration"部分 - 明确 SKILL 提供参考材料用于智能整合，而非直接复制模板 - 定义响应策略：分析上下文 → 提取相关信息 → 合成定制化 → 交付目标响应 - 列出禁止事项和必须遵守的原则 2. 更新所有 6 个 Mode 的 Process 描述： - Mode 1 (Command Search): 强调智能过滤和排序，而非 JSON 数据转储 - Mode 2 (Smart Recommendations): 分析工作流上下文并评估推荐优先级 - Mode 3 (Full Documentation): 提取用户相关部分，渐进式披露 - Mode 4 (Beginner Onboarding): 评估用户背景，设计个性化学习路径 - Mode 5 (Issue Reporting): 智能引导信息收集，适应模板部分 - Mode 6 (Deep Command Analysis): 合成聚焦解释，处理并整合 CLI 分析 3. 更新示例说明： - 每个 Mode 的示例都强调"NOT: [直接返回模板]" - 展示如何根据上下文定制响应 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude --- .claude/skills/command-guide/SKILL.md | 119 ++++++++++++++++++-------- 1 file changed, 84 insertions(+), 35 deletions(-) diff --git a/.claude/skills/command-guide/SKILL.md b/.claude/skills/command-guide/SKILL.md index f6e1a42a..be837609 100644 --- a/.claude/skills/command-guide/SKILL.md +++ b/.claude/skills/command-guide/SKILL.md @@ -8,6 +8,29 @@ allowed-tools: Read, Grep, Glob, AskUserQuestion Comprehensive command guide for Claude DMS3 workflow system covering 69 commands across 4 categories (workflow, cli, memory, task). +## 🧠 Core Principle: Intelligent Integration + +**⚠️ IMPORTANT**: This SKILL provides **reference materials** for intelligent integration, NOT templates for direct copying. + +**Response Strategy**: +1. **Analyze user's specific context** - Understand their exact need, workflow stage, and technical level +2. **Extract relevant information** - Select only the pertinent parts from reference guides +3. **Synthesize and customize** - Combine multiple sources, add context-specific examples +4. **Deliver targeted response** - Provide concise, actionable guidance tailored to the user's situation + +**Never**: +- ❌ Copy-paste entire template sections verbatim +- ❌ Return raw reference documentation without processing +- ❌ Provide generic responses that ignore user context + +**Always**: +- ✅ Understand the user's specific situation first +- ✅ Integrate information from multiple sources (indexes, guides, reference docs) +- ✅ Customize examples and explanations to match user's use case +- ✅ Provide progressive depth - brief answers with "more detail available" prompts + +--- + ## 🎯 Operation Modes ### Mode 1: Command Search 🔍 @@ -19,10 +42,11 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands **Process**: 1. Identify search type (keyword/category/use-case) 2. Query appropriate index (all-commands/by-category/by-use-case) -3. Return matching commands with metadata -4. Suggest related commands +3. **Intelligently filter and rank** results based on user's implied context +4. **Synthesize concise response** with command names, brief descriptions, and use-case fit +5. **Suggest next steps** - related commands or workflow patterns -**Example**: "搜索 planning 命令" → Lists planning commands from `index/by-use-case.json` +**Example**: "搜索 planning 命令" → Analyze user's likely goal → Present top 3-5 most relevant planning commands with context-specific usage hints, NOT raw JSON dump --- @@ -33,12 +57,16 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands **Triggers**: "下一步", "what's next", "after /workflow:plan", "推荐" **Process**: -1. Parse current context (last command/workflow state) -2. Query `index/command-relationships.json` -3. Return recommended next commands with rationale -4. Show common workflow patterns +1. **Analyze workflow context** - Understand where user is in their development cycle +2. Query `index/command-relationships.json` for possible next commands +3. **Evaluate and prioritize** recommendations based on: + - User's stated goals + - Common workflow patterns + - Project complexity indicators +4. **Craft contextual guidance** - Explain WHY each recommendation fits, not just WHAT to run +5. **Provide workflow examples** - Show complete flow, not isolated commands -**Example**: "执行完 /workflow:plan 后做什么？" → Recommends /workflow:execute or /workflow:action-plan-verify +**Example**: "执行完 /workflow:plan 后做什么？" → Analyze plan output quality → Recommend `/workflow:action-plan-verify` (if complex) OR `/workflow:execute` (if straightforward) with reasoning for each choice --- @@ -51,10 +79,11 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands **Process**: 1. Locate command in `index/all-commands.json` 2. Read original command file for full details -3. Present parameters, arguments, examples -4. Link to related commands +3. **Extract user-relevant sections** - Focus on what they asked about (parameters OR examples OR workflow) +4. **Enhance with context** - Add use-case specific examples if user's scenario is clear +5. **Progressive disclosure** - Provide core info first, offer "need more details?" prompts -**Example**: "/workflow:plan 的参数是什么？" → Shows full parameter list and usage examples +**Example**: "/workflow:plan 的参数是什么？" → Identify user's experience level → Present parameters with context-appropriate explanations (beginner: verbose + examples; advanced: concise + edge cases), NOT raw documentation dump --- @@ -65,12 +94,13 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands **Triggers**: "新手", "getting started", "如何开始", "常用命令" **Process**: -1. Present progressive learning path -2. Show `index/essential-commands.json` (Top 14 commands) -3. Link to getting-started guide -4. Provide first workflow example +1. **Assess user background** - Ask clarifying questions if needed (coding experience? project type?) +2. **Design personalized learning path** based on their goals +3. **Curate essential commands** from `index/essential-commands.json` - Select 3-5 most relevant for their use case +4. **Provide guided first example** - Walk through ONE complete workflow with explanation +5. **Set clear next steps** - What to try next, where to get help -**Example**: "我是新手，如何开始？" → Learning path + Top 14 commands + quick start guide +**Example**: "我是新手，如何开始？" → Detect if they have a specific task OR just exploring → For specific task: provide laser-focused 3-step guide; For exploring: progressive learning path starting with simplest workflow, NOT overwhelming 14-command list --- @@ -81,17 +111,21 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands **Triggers**: **"CCW-issue"**, **"CCW-help"**, **"ccw-issue"**, **"ccw-help"**, **"ccw"**, "报告 bug", "功能建议", "问题咨询", "交互式诊断" **Process**: -1. Use AskUserQuestion to confirm type (diagnosis/bug/feature/question) -2. Collect required information interactively with **execution flow emphasis** -3. Select appropriate template: - - `issue-diagnosis.md` - Full diagnostic workflow with decision tree - - `issue-bug.md` - Bug report with complete command history - - `issue-feature.md` - Feature request with current workflow analysis - - `issue-question.md` - Question with detailed attempt history -4. Generate filled template with privacy-protected command history -5. Save/display with troubleshooting guidance +1. **Understand issue context** - Use AskUserQuestion to confirm type AND gather initial context +2. **Intelligently guide information collection**: + - Adapt questions based on previous answers + - Skip irrelevant sections + - Probe for missing critical details +3. **Select and customize template**: + - `issue-diagnosis.md`, `issue-bug.md`, `issue-feature.md`, or `issue-question.md` + - **Adapt template sections** to match user's specific scenario +4. **Synthesize coherent issue report**: + - Integrate collected information with appropriate template sections + - **Highlight key details** - Don't bury critical info in boilerplate + - Add privacy-protected command history +5. **Provide actionable next steps** - Immediate troubleshooting OR submission guidance -**Example**: "CCW-issue" → Interactive Q&A → Generates GitHub issue template with full execution context +**Example**: "CCW-issue" → Detect user frustration level → For urgent: fast-track to critical info collection; For exploratory: comprehensive diagnostic flow, NOT one-size-fits-all questionnaire **🆕 Enhanced Features**: - Complete command history with privacy protection @@ -116,12 +150,19 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands **Simple Query** (direct documentation lookup): 1. Identify target command/agent from user query 2. Locate corresponding markdown file in `reference/` -3. Read and extract relevant sections -4. Present formatted response with examples +3. **Extract contextually relevant sections** - Not entire document +4. **Synthesize focused explanation**: + - Address user's specific question + - Add context-appropriate examples + - Link related concepts +5. **Offer progressive depth** - "Want to know more about X?" **Complex Query** (CLI-assisted analysis): -1. Detect complexity indicators (多个命令对比、工作流程分析、最佳实践) -2. Construct analysis prompt for gemini/qwen: +1. **Detect complexity indicators** (多个命令对比、工作流程分析、最佳实践) +2. **Design targeted analysis prompt** for gemini/qwen: + - Frame user's question precisely + - Specify required analysis depth + - Request structured comparison/synthesis ```bash gemini -p " PURPOSE: Analyze command documentation to answer user query @@ -133,7 +174,11 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands " -m gemini-3-pro-preview-11-2025 --include-directories ~/.claude/skills/command-guide/reference ``` Note: Use absolute path `~/.claude/skills/command-guide/reference` for reference documentation access -3. Return CLI analysis results to user +3. **Process and integrate CLI analysis**: + - Extract key insights from CLI output + - Add context-specific examples + - Synthesize actionable recommendations +4. **Deliver tailored response** - Not raw CLI output **Query Classification**: - **Simple**: Single command explanation, parameter list, basic usage @@ -145,16 +190,20 @@ Comprehensive command guide for Claude DMS3 workflow system covering 69 commands ``` User: "action-planning-agent 如何工作？" → Read reference/agents/action-planning-agent.md -→ Extract workflow, capabilities, examples -→ Present structured response +→ **Identify user's knowledge gap** (mechanism? inputs/outputs? when to use?) +→ **Extract relevant sections** addressing their need +→ **Synthesize focused explanation** with examples +→ NOT: Dump entire agent documentation ``` *Complex Query*: ``` User: "对比 workflow:plan 和 workflow:tdd-plan 的使用场景和最佳实践" → Detect: 多命令对比 + 最佳实践 -→ Use gemini analysis on reference/commands/workflow/ -→ Return comprehensive comparison with workflow examples +→ **Design comparison framework** (when to use, trade-offs, workflow integration) +→ Use gemini to analyze both commands with structured comparison prompt +→ **Synthesize insights** into decision matrix and usage guidelines +→ NOT: Raw command documentation side-by-side ``` ---