research-topic

借助代码搜索、记忆存储、知识库，以及可选的网络搜索，对某一主题进行深入研究。当您需要在采取行动之前充分理解某个问题时，可使用此功能。

SKILL.md

--- frontmatter

name: research-topic
description: Research a topic thoroughly using code search, memory, knowledge, and optional web search. Use when you need to understand something before taking action.

Research Topic

Deep dive on a topic using internal and external sources. Searches codebase, memory, project knowledge, and optionally the web.

Inputs

Input	Type	Default	Purpose
`topic`	string	required	Topic to research (e.g., "pytest fixtures", "Redis caching")
`project`	string	auto	Project to search (from config)
`depth`	string	"normal"	"quick" (code only), "normal" (code + memory), "deep" (all sources)
`focus`	string	-	Specific aspect (e.g., "performance", "security")

Workflow

1. Detect Project

•If project empty: infer from cwd via config repositories; default automation-analytics-backend

2. Search Codebase

•code_search(query=topic + focus, project=project, limit=10)
•Parse results into file paths, scores, previews

3. Check Memory (depth != "quick")

•memory_read("learned/patterns")
•Filter patterns relevant to topic/focus

4. Query Project Knowledge (depth != "quick")

•knowledge_query(project=project, section="architecture")
•knowledge_query(project=project, section="patterns")
•knowledge_query(project=project, section="gotchas")

5. Optional External Search

•For "deep" or when internal sources insufficient: WebSearch for external docs

6. Build Summary

Output markdown with:

•Code Found: file list with previews
•Past Learnings: relevant patterns from memory
•Architecture Context: from knowledge
•Coding Patterns and Gotchas
•Next Steps: explain_code, plan_implementation, persona_load("developer")

7. Log

•memory_session_log("Researched: {topic}", "Project: {project}, Depth: {depth}")

Key MCP Tools

•code_search — semantic code search
•memory_read — learned patterns
•knowledge_query — architecture, patterns, gotchas
•memory_session_log — session logging
•WebSearch — external documentation (optional)