Refactor Skill for Claude
Code refactoring workflow - analyze → plan → implement → review → validate
Install
one command, takes just this skill from the repositorynpx skills add https://github.com/parcadei/Continuous-Claude-v3 --skill refactorThe instruction itself
17 sections, as written by the author/refactor - Refactoring Workflow
Safe refactoring with review gates.
When to Use
- "Refactor X"
- "Clean up this code"
- "Extract this into a module"
- "Improve the architecture of Y"
- Large-scale code restructuring
- Technical debt reduction
Workflow Overview
┌──────────┐ ┌────────────┐ ┌──────────┐ ┌──────────┐ ┌───────────┐
│ phoenix │───▶│ plan- │───▶│ kraken │───▶│plan-reviewer│───▶│ arbiter │
│ │ │ agent │ │ │ │ │ │ │
└──────────┘ └────────────┘ └──────────┘ └──────────┘ └───────────┘
Analyze Plan Implement Review Verify
current changes refactor changes tests pass
Agent Sequence
| # | Agent | Role | Output |
|---|-------|------|--------|
| 1 | phoenix | Analyze current code, identify improvement areas | Analysis report |
| 2 | plan-agent | Create safe refactoring plan | Step-by-step plan |
| 3 | kraken | Implement the refactoring | Code changes |
| 4 | plan-reviewer | Review changes for correctness | Review report |
| 5 | arbiter | Verify all tests still pass | Test report |
Refactoring Principles
- Tests first: Ensure adequate test coverage before refactoring
- Small steps: Each change should be independently verifiable
- Behavior preserved: No functional changes during refactor
- Reviewable: Changes should be easy to review
Execution
Phase 1: Analyze
Task(
subagent_type="phoenix",
prompt="""
Analyze for refactoring: [TARGET_CODE]
Identify:
- Current pain points
- Code smells
- Improvement opportunities
- Risk areas
- Test coverage gaps
"""
)
Phase 2: Plan
Task(
subagent_type="plan-agent",
prompt="""
Plan refactoring: [TARGET_CODE]
Analysis: [from phoenix]
Create:
- Step-by-step refactoring plan
- Each step should be:
- Small and focused
- Independently testable
- Reversible
- Identify files affected
- Risk mitigation strategy
"""
)
Phase 3: Implement
Task(
subagent_type="kraken",
prompt="""
Implement refactoring: [TARGET_CODE]
Plan: [from plan-agent]
Requirements:
- Follow plan exactly
- Run tests after each step
- Stop if tests fail
- NO behavior changes
"""
)
Phase 4: Review
Task(
subagent_type="plan-reviewer",
prompt="""
Review refactoring: [TARGET_CODE]
Changes: [git diff from kraken]
Check:
- Behavior preserved
- No unintended changes
- Code quality improved
- Patterns consistent
"""
)
Phase 5: Validate
Task(
subagent_type="arbiter",
prompt="""
Validate refactoring: [TARGET_CODE]
- Run full test suite
- Verify no regressions
- Check type errors
- Run linting
"""
)
Refactoring Types
Extract Module
phoenix → plan-agent → kraken → plan-reviewer → arbiter
Rename/Restructure
phoenix → kraken → arbiter (simpler, skip detailed planning)
Architecture Change
phoenix → plan-agent → [kraken → plan-reviewer] × N phases → arbiter
Example
User: /refactor Extract the validation logic into a separate module
Claude: Starting /refactor workflow...
Phase 1: Analyzing current structure...
[Spawns phoenix]
Found: Validation logic spread across 4 files
- form.ts (lines 45-120)
- api.ts (lines 200-280)
- user.ts (lines 15-45)
- order.ts (lines 88-130)
Phase 2: Planning extraction...
[Spawns plan-agent]
Plan:
1. Create src/validation/index.ts
2. Extract common validators
3. Update imports one file at a time
4. Run tests after each change
Phase 3: Implementing...
[Spawns kraken]
Completed all 4 steps, tests green after each
Phase 4: Reviewing changes...
[Spawns plan-reviewer]
✅ All behavior preserved
✅ DRY improved (removed 45 duplicate lines)
✅ New structure consistent
Phase 5: Final validation...
[Spawns arbiter]
✅ 312 tests passing, 0 regressions
Refactoring complete!
Safety Flags
--dry-run: Plan but don't implement--step-by-step: Pause after each change for approval--coverage-check: Require >80% coverage before proceeding
Other skills for the same job
different authors, same section of the catalogue
Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
Execute git commit with conventional commit message analysis, intelligent staging, and message generation. Use when user asks to commit changes, create a git commit, or mentions "/commit". Supports: (1) Auto-detecting type and scope from changes, (2) Generating conventional commit messages from diff, (3) Interactive commit with optional type/scope/description overrides, (4) Intelligent file staging for logical grouping
Comprehensive GitHub code review with AI-powered swarm coordination
Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.
Use this skill to review code. It supports both local changes (staged or working tree) and remote Pull Requests (by ID or URL). It focuses on correctness, maintainability, and adherence to project standards.
Refactor bloated AGENTS.md, CLAUDE.md, or similar agent instruction files to follow progressive disclosure principles. Splits monolithic files into organized, linked documentation.
Create high-quality git commits: review/stage intended changes, split into logical commits, and write clear commit messages (including Conventional Commits). Use when the user asks to commit, craft a commit message, stage changes, or split work into multiple commits.
How to use it
Take parcadei/refactor from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.