ワンクリックで
refactor
Analyze code for structural improvement opportunities and produce a prioritized refactoring plan
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
Analyze code for structural improvement opportunities and produce a prioritized refactoring plan
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
Run parallel code review agents and consolidate findings into a unified report
Capture a follow-up idea or task as a work-item draft without interrupting current work. Use to quickly jot down something you think of mid-task so it gets tracked. Works with nibs, GitHub, Azure DevOps, or a Markdown fallback.
Take an under-specified work item and make it actionable — resolve its open questions through a short interview grounded in the code, and give it acceptance criteria. Use on a captured draft, or any item too vague to start on.
Reconcile what was built against what was planned, record decisions and deviations, close the item (a single phase or a whole plan), and file follow-ups for deferred work. Use after finishing a phase or plan to keep planned and actual from drifting apart.
Orchestrate execution of MULTIPLE nibs in one run. Selects a queue, understands the nibs collectively (including how they fit together), then chooses the best execution mechanism per cluster — single series agent, parallel fan-out, scripted workflow, or agent team — and dispatches with ONE approval gate. Use when the user wants to work several nibs together (in parallel or series) rather than one at a time. Complements /decaf-build:auto-dev and /decaf-build:auto-tdd (which handle a single nib).
Direct development with automated review. Plans implementation, executes via subagent, then auto-reviews. Use for work that isn't test-driven (UI, config, styling, infrastructure, scaffolding).
| name | refactor |
| description | Analyze code for structural improvement opportunities and produce a prioritized refactoring plan |
| argument-hint | [quick|deep] [path|glob|full] [instructions] |
This command analyzes existing code for structural improvement opportunities and produces a prioritized, actionable refactoring plan. It focuses on high-impact opportunities, not exhaustive issue lists. The output is a local plan; walk and act on it with /decaf-quality:resolve-refactor.
Parse $ARGUMENTS to determine:
quick or deep (default when no mode keyword given)full, or changed files (default)| Mode | Agents | Use Case |
|---|---|---|
quick | structural-analyst only (1) | Fast local quality check on specific files |
deep (default) | structural-analyst + coherence-analyst (2) | Full analysis including cross-cutting patterns |
| Input | Behavior |
|---|---|
| (none) | Changed files from git diff HEAD --name-only. Fall back to git diff HEAD~1..HEAD --name-only if no uncommitted changes. |
src/Services/ | All source files in that directory (recursive) |
src/**/*.cs | Files matching the glob pattern |
src/OrderProcessor.cs | That single file |
full | Entire project — smart-sampled (see below) |
full ScopeWhen scope is full, the codebase may be too large to analyze entirely. Apply smart sampling:
To determine exclusions, check CLAUDE.md for generated/vendored file patterns. Use .gitignore as baseline.
Determine which files to analyze based on scope resolution above.
# For default scope (changed files):
git diff HEAD --name-only
# For directory scope:
# Use Glob tool to find source files
# For full scope:
# Use Glob + smart sampling
If scope is empty (no files match), inform the user and exit:
No files found matching the specified scope. Provide a file path, directory, or glob pattern.
Read the project's CLAUDE.md to understand:
quick mode1 agent — no cross-file analysis:
decaf-quality:structural-analystdeep mode (default)2 agents in parallel:
decaf-quality:structural-analystdecaf-quality:coherence-analystCRITICAL: In deep mode, both agents MUST be launched in a single message with multiple Agent tool calls. This ensures true parallel execution.
structural-analyst prompt:
Analyze the following files for per-file structural improvement opportunities.
Focus on your area of expertise. Follow your own output format instructions.
## Files to Analyze
<list each file path, and paste file contents>
## Project Context
<relevant CLAUDE.md excerpts — language, conventions, architecture>
## Additional Instructions
<any user-provided instructions from $ARGUMENTS>
coherence-analyst prompt (deep mode only):
Analyze the following files for cross-file structural patterns and improvement opportunities.
Focus on your area of expertise. Follow your own output format instructions.
## Files to Analyze
<list each file path>
## Project Context
<relevant CLAUDE.md excerpts — language, conventions, architecture, module structure>
## Additional Instructions
<any user-provided instructions from $ARGUMENTS>
Note: You have access to Grep and Glob tools to search the broader codebase
for patterns beyond the listed files. Use them to verify cross-file patterns.
Wait for all agents to complete. Each agent returns opportunities in JSON format.
Apply the consolidation rules:
@../../conventions/refactoring.md
For each agent's "Considered But Not Flagged" section:
Create a timestamped plan file in .decaf/refactoring-plans/ at the repo root. Never overwrite existing plans.
Filename: .decaf/refactoring-plans/REFACTOR_PLAN_<YYYY-MM-DD>_<HH-MM-SS>.md using the current date and time you know from context. Do NOT shell out to date — construct the filename directly. Create the .decaf/refactoring-plans/ directory first if it doesn't exist.
Plan format:
# Refactoring Plan
**Mode**: <mode> | **Analysts**: <agent list> | **Date**: <YYYY-MM-DD>
**Scope**: N files analyzed
## Value Matrix
| Rating | Count |
|--------|-------|
| ★★★ | X |
| ★★ | X |
**Total opportunities:** N (grouped into M refactoring units)
---
## Refactoring Units
### #1 ★★★ Extract PaymentValidation domain service
| | |
|---|---|
| **Impact** | High |
| **Effort** | Medium |
| **Files** | `src/OrderProcessor.cs`, `src/PaymentService.cs`, `src/RefundHandler.cs` |
| **Category** | validation-scattering |
| **Found by** | coherence-analyst (High), structural-analyst (Medium) |
| **Confidence** | 88/100 |
**Problem:** Payment validation logic is duplicated across 3 services with subtly different implementations. The OrderProcessor validates card expiry differently than PaymentService, creating inconsistent behavior.
**Before:**
```csharp
// OrderProcessor.cs:45
if (card.ExpiryDate < DateTime.Now) ...
// PaymentService.cs:72
if (card.ExpiryDate <= DateTime.UtcNow) ...
After:
// PaymentValidation.cs
public static bool IsCardExpired(Card card)
=> card.ExpiryDate <= DateTime.UtcNow;
Steps:
PaymentValidation service with consolidated rules| Impact | High |
| Effort | Small |
| Files | src/OrderProcessor.cs |
| Category | god-function |
| Found by | structural-analyst |
| Confidence | 92/100 |
Problem: ProcessOrder() is 85 lines mixing validation, persistence, and notification concerns at different abstraction levels.
Before:
public async Task ProcessOrder(Order order)
{
// 85 lines mixing validation, DB calls, email sending
}
After:
public async Task ProcessOrder(Order order)
{
Validate(order);
await Persist(order);
await NotifyCustomer(order);
}
Steps:
...
[Items examined but below inclusion threshold or determined to be acceptable design, grouped by agent with rationale.]
Scope: [Description of what was analyzed] Sampling: [If full mode, note which files were sampled/excluded] Limitations: [What cross-cutting patterns may have been missed due to scope]
### Severity Icons for Star Ratings
- ★★★ — High-value: address first
- ★★ — Worthwhile: address when nearby
- ★ — Low-value (not included in main plan)
**Always use literal Unicode star characters (★), never `:shortcode:` syntax.**
### Output Notification
After creating the plan file, inform the user:
✅ Refactoring plan complete: .decaf/refactoring-plans/REFACTOR_PLAN_2026-03-03_14-30-45.md
★★★ High-value: X opportunities ★★ Worthwhile: Y opportunities Total: Z refactoring units
Run /decaf-quality:resolve-refactor to walk through opportunities interactively.
## Example Usage
/decaf-quality:refactor # deep mode, changed files /decaf-quality:refactor quick # quick mode (structural only) /decaf-quality:refactor src/Services/ # deep mode, specific directory /decaf-quality:refactor src/**/*.cs # deep mode, glob pattern /decaf-quality:refactor src/OrderProcessor.cs # deep mode, single file /decaf-quality:refactor full # deep mode, entire project (sampled) /decaf-quality:refactor quick src/OrderProcessor.cs # quick mode, single file /decaf-quality:refactor focus on error handling # deep mode with instructions