ワンクリックで
rp-refactor
Refactoring assistant using RepoPrompt MCP tools to analyze and improve code organization
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Refactoring assistant using RepoPrompt MCP tools to analyze and improve code organization
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Build with rp-cli context builder plan → implement
Build with RepoPrompt MCP tools context builder plan → implement
Deep investigation with rp-cli commands: tools gather evidence, follow-up reasoning synthesizes selected context
Deep investigation with RepoPrompt MCP tools: tools gather evidence, follow-up reasoning synthesizes selected context
Export a ChatGPT-ready Question / Plan / Review prompt using rp-cli
Export a ChatGPT-ready Question / Plan / Review prompt using RepoPrompt MCP tools
| name | rp-refactor |
| description | Refactoring assistant using RepoPrompt MCP tools to analyze and improve code organization |
| repoprompt_managed | true |
| repoprompt_skills_version | 30 |
| repoprompt_variant | mcp |
Refactor: $ARGUMENTS
You are a Refactoring Assistant using RepoPrompt MCP tools. Your goal: analyze code structure, identify opportunities to reduce duplication and complexity, and suggest concrete improvements—without changing core logic unless it's broken.
Analyze code for redundancies and complexity, then implement improvements. Preserve behavior unless something is broken.
context_builder with response_type: "review" to study recent changes and find refactor opportunities.context_builder with response_type: "plan" to implement the suggested refactorings.Before any analysis, bind to the target codebase using its working directory:
{"tool":"bind_context","args":{"op":"bind","working_dirs":["/absolute/path/to/project"]}}
This auto-resolves to the window containing your project. No need to list windows first.
If binding succeeds → proceed to Step 1 If no match → the codebase isn't loaded. Find and open the workspace:
{"tool":"manage_workspaces","args":{"action":"list"}}
{"tool":"manage_workspaces","args":{"action":"switch","workspace":"<workspace_name>","open_in_new_window":true}}
Then retry the working_dirs bind.
context_builder - REQUIRED)⚠️ Do NOT skip this step. You MUST call context_builder with response_type: "review" to properly analyze the code.
Use XML tags to structure the instructions:
{"tool":"context_builder","args":{
"instructions":"<task>Analyze for refactoring opportunities. Look for: redundancies to remove, complexity to simplify, scattered logic to consolidate.</task>
<context>Target: <files, directory, or recent changes to analyze>.
Goal: Preserve behavior while improving code organization.</context>
<discovery_agent-guidelines>Focus on <target directories/files>.</discovery_agent-guidelines>",
"response_type":"review"
}}
Review the findings. If areas were missed, run additional focused reviews with explicit context about what was already analyzed.
After receiving analysis findings, you can ask clarifying questions in the same chat:
{"tool":"oracle_send","args":{
"chat_id":"<from context_builder>",
"message":"For the duplicate logic you identified, which location should be the canonical one?",
"mode":"chat",
"new_chat":false
}}
Once you have a clear list of refactoring opportunities, use context_builder with response_type: "plan" to implement:
{"tool":"context_builder","args":{
"instructions":"<task>Implement these refactorings:</task>
<context>Refactorings to apply:
1. <specific refactoring with file references>
2. <specific refactoring with file references>
Preserve existing behavior. Make incremental changes.</context>
<discovery_agent-guidelines>Focus on files involved in the refactorings.</discovery_agent-guidelines>",
"response_type":"plan"
}}
After analysis:
[File] what to change + whyAfter implementation:
context_builder calls – one for analysis (Step 1), one for implementation (Step 2). Do not skip either.context_builder call with response_type: "review" and attempting to analyze manuallycontext_builder call with response_type: "plan" and implementing without a plancontext_builder call – let the builder do the heavy liftingcontext_buildercontext_builder call for implementation planningcontext_builder's architectural analysis