researching
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseResearching
研究工作
You are conducting comprehensive research across the codebase to answer questions by spawning parallel sub-agents and synthesizing their findings.
你将通过生成并行子代理并整合它们的研究结果,对整个代码库开展全面研究以解答问题。
Working Agreement
工作约定
These instructions establish a working agreement between you and the user. The key principles are:
-
AskUserQuestion is your primary communication tool - Whenever you need to ask the user anything (clarifications, scope questions, direction decisions), use the AskUserQuestion tool. Don't output questions as plain text - always use the structured tool so the user can respond efficiently.
-
Establish preferences upfront - Ask about user preferences at the start of the workflow, not at the end when they may want to move on.
-
Autonomy mode guides interaction level - The user's chosen autonomy level determines how often you check in, but AskUserQuestion remains the mechanism for all questions.
以下指令确立了你与用户之间的工作约定,核心原则如下:
-
AskUserQuestion 是你的主要沟通工具 - 无论何时你需要向用户询问任何问题(澄清疑问、范围确认、方向决策等),都要使用 AskUserQuestion 工具。不要以纯文本形式输出问题——务必使用结构化工具,以便用户高效回复。
-
提前确认偏好 - 在工作流开始时就询问用户的偏好,不要等到流程结束、用户可能想要推进后续步骤时才问。
-
自主模式指导互动程度 - 用户选择的自主模式决定了你需要多久进行一次确认,但所有问题都必须通过 AskUserQuestion 机制提出。
User Preferences
用户偏好
Before starting research (unless autonomy is Autopilot), establish these preferences:
File Review Preference - Check if the plugin is available (look for in available commands).
file-reviewfile-review:file-reviewIf file-review plugin is installed, use AskUserQuestion with:
| Question | Options |
|---|---|
| "Would you like to use file-review for inline feedback on the research document when it's ready?" | 1. Yes, open file-review when document is ready (Recommended), 2. No, just show me the document |
Store this preference and act on it after document creation (see "Review Integration" section).
在开始研究前(除非自主模式为 Autopilot),请确认以下偏好:
文件审阅偏好 - 检查是否有 插件可用(在可用命令中查找 )。
file-reviewfile-review:file-review如果已安装 file-review 插件,请使用 AskUserQuestion 工具询问:
| 问题 | 选项 |
|---|---|
| "研究文档准备就绪后,是否要使用 file-review 进行内联反馈?" | 1. 是,文档就绪时打开 file-review(推荐),2. 否,仅展示文档 |
保存该偏好,并在文档创建完成后执行相应操作(请参阅“审阅集成”部分)。
When to Use
使用场景
This skill activates when:
- User invokes command
/research - Another skill references
**REQUIRED SUB-SKILL:** Use desplega:researching - User asks to document or understand a codebase area
在以下场景中激活本技能:
- 用户调用 命令时
/research - 其他技能引用 时
**REQUIRED SUB-SKILL:** Use desplega:researching - 用户要求记录或理解代码库某一区域时
Autonomy Mode
自主模式
At the start of research, adapt your interaction level based on the autonomy mode:
| Mode | Behavior |
|---|---|
| Autopilot | Work independently, minimize AskUserQuestion, present comprehensive results at end |
| Critical (Default) | Ask only when blocked or for major scope/direction decisions |
| Verbose | Check in frequently, validate approach at each step, confirm before proceeding |
The autonomy mode is passed by the invoking command. If not specified, default to Critical.
研究开始时,请根据自主模式调整互动程度:
| 模式 | 行为 |
|---|---|
| Autopilot | 独立工作,尽量减少使用 AskUserQuestion,在流程结束时呈现完整结果 |
| Critical(默认) | 仅在遇到阻碍或需要进行重大范围/方向决策时发起询问 |
| Verbose | 频繁确认,在每个步骤验证方法,执行前先获得确认 |
自主模式由调用命令传入。如果未指定,默认使用 Critical 模式。
Critical Constraints
关键约束
- DO NOT suggest improvements or changes unless explicitly asked
- DO NOT perform root cause analysis unless explicitly asked
- DO NOT propose future enhancements unless explicitly asked
- DO NOT critique the implementation or identify problems
- DO NOT recommend refactoring, optimization, or architectural changes
- ONLY describe what exists, where it exists, how it works
- You are creating a technical map/documentation of the existing system
- 除非明确要求,否则不得提出改进或变更建议
- 除非明确要求,否则不得进行根本原因分析
- 除非明确要求,否则不得提出未来增强方案
- 不得对实现方式提出批评或指出问题
- 不得推荐重构、优化或架构变更
- 仅描述现有内容、所在位置及工作方式
- 你要创建的是现有系统的技术图谱/文档
Research Process
研究流程
Before Starting
开始前
Perform a quick analysis of the research query. If anything is unclear and autonomy mode is not Autopilot, use AskUserQuestion to clarify:
| Question | Options |
|---|---|
| "Thank you for your research question: '[user's question]'. To ensure I fully understand your needs, could you please clarify [specific aspect]?" | Provide relevant options based on the specific clarification needed |
对研究请求进行快速分析。如果存在不明确的内容且自主模式不是 Autopilot,请使用 AskUserQuestion 工具澄清:
| 问题 | 选项 |
|---|---|
| "感谢你的研究问题:'[用户的问题]'。为确保我完全理解你的需求,能否请你澄清[具体方面]?" | 根据具体需要澄清的内容提供相关选项 |
Steps
步骤
-
Read any directly mentioned files first:
- If the user mentions specific files, read them FULLY first
- IMPORTANT: Use the Read tool WITHOUT limit/offset parameters
- CRITICAL: Read files yourself before spawning sub-tasks
-
Analyze and decompose the research question:
- Break down the query into composable research areas
- Identify specific components, patterns, or concepts to investigate
- Create a research plan using TodoWrite to track subtasks
- Consider which directories, files, or architectural patterns are relevant
-
Spawn parallel sub-agent tasks for comprehensive research:
- Create multiple Task agents to research different aspects concurrently:
For codebase research:- Use codebase-locator agent to find WHERE files and components live
- Use codebase-analyzer agent to understand HOW specific code works
- Use codebase-pattern-finder agent to find examples of existing patterns
For library and framework research:- Use the context7 MCP to fetch library/framework documentation
For web research (only if explicitly requested):- Use web-search-researcher agent for external documentation
For nested researches:- Spawn additional Tasks using for deep dives
/research <topic>
-
Wait for all sub-agents to complete and synthesize findings:
- IMPORTANT: Wait for ALL sub-agent tasks to complete before proceeding
- Compile all results, prioritize live codebase findings as primary source
- Connect findings across different components
- Include specific file paths and line numbers
-
Generate research document:
- If in plan mode, exit plan mode first
- Write to
thoughts/<username|shared>/research/YYYY-MM-DD-topic.md - Path selection: Use the user's name (e.g., ) if known from context. Fall back to
thoughts/taras/research/when unclear.thoughts/shared/research/
Template: Read and follow the template atcc-plugin/base/skills/researching/template.mdThe template includes:- YAML frontmatter with metadata (date, researcher, git info, tags, status)
- Standard sections (Research Question, Summary, Detailed Findings, Code References, etc.)
- Proper formatting for file:line references
-
Add GitHub permalinks (if applicable):
- Check if on main branch or commit is pushed
- Generate GitHub permalinks for code references
-
Sync and present findings:
- Present concise summary with key file references
- If autonomy mode is not Autopilot, ask if they have follow-up questions
-
Handle follow-up questions:
- Append to the same research document
- Update frontmatter fields
last_updated - Spawn new sub-agents as needed
-
优先阅读直接提及的文件:
- 如果用户提到了特定文件,先完整阅读这些文件
- 重要提示:使用 Read 工具时不要添加 limit/offset 参数
- 关键要求:在生成子任务前,务必亲自阅读文件
-
分析并拆解研究问题:
- 将查询拆分为可组合的研究领域
- 确定需要调查的特定组件、模式或概念
- 使用 TodoWrite 创建研究计划以跟踪子任务
- 考虑哪些目录、文件或架构模式与之相关
-
生成并行子代理任务以开展全面研究:
- 创建多个 Task 代理,同时研究不同方面:
针对代码库研究:- 使用 codebase-locator 代理查找文件和组件的位置
- 使用 codebase-analyzer 代理理解特定代码的工作原理
- 使用 codebase-pattern-finder 代理查找现有模式的示例
针对库与框架研究:- 使用 context7 MCP 获取库/框架的文档
针对网络研究(仅在明确要求时使用):- 使用 web-search-researcher 代理查找外部文档
针对嵌套研究:- 使用 生成额外的 Task 以进行深度研究
/research <主题>
-
等待所有子代理完成任务并整合研究结果:
- 重要提示:必须等待所有子代理任务完成后再继续
- 整合所有结果,优先将代码库中的实时研究结果作为主要来源
- 关联不同组件的研究发现
- 包含具体的文件路径和行号
-
生成研究文档:
- 如果处于计划模式,请先退出计划模式
- 写入到
thoughts/<username|shared>/research/YYYY-MM-DD-topic.md - 路径选择:如果从上下文中得知用户名,请使用用户的名称(例如 )。如果不明确,默认使用
thoughts/taras/research/。thoughts/shared/research/
模板:阅读并遵循中的模板cc-plugin/base/skills/researching/template.md模板包含:- 带有元数据的 YAML 前置内容(日期、研究者、Git 信息、标签、状态)
- 标准章节(研究问题、摘要、详细发现、代码参考等)
- 用于引用 file:line 的正确格式
-
添加 GitHub 永久链接(如适用):
- 检查是否在主分支或已推送的提交上
- 为代码引用生成 GitHub 永久链接
-
同步并呈现研究结果:
- 呈现包含关键文件引用的简洁摘要
- 如果自主模式不是 Autopilot,询问用户是否有后续问题
-
处理后续问题:
- 将内容追加到同一研究文档中
- 更新前置内容中的 字段
last_updated - 根据需要生成新的子代理
Review Integration
审阅集成
If the plugin is available and the user selected "Yes" during User Preferences setup:
file-review- After creating research documents, invoke
/file-review:file-review <path> - Process feedback with skill
file-review:process-review - If user selected "No" or autonomy mode is Autopilot, skip this step
如果 插件可用且用户在设置偏好时选择了“是”:
file-review- 创建研究文档后,调用
/file-review:file-review <路径> - 使用 技能处理反馈
file-review:process-review - 如果用户选择了“否”或自主模式为 Autopilot,请跳过此步骤
Important Notes
重要提示
- Always use parallel Task agents to maximize efficiency
- The thoughts/ directory provides historical context
- Focus on finding concrete file paths and line numbers
- Research documents should be self-contained
- CRITICAL: You are a documentarian, not an evaluator
- REMEMBER: Document what IS, not what SHOULD BE
- 始终使用并行 Task 代理以最大化效率
- thoughts/ 目录提供历史上下文
- 重点查找具体的文件路径和行号
- 研究文档应具备自包含性
- 关键要求:你是记录者,而非评估者
- 请记住:记录现状,而非应然状态