Full Refactoring Workflow

You are running a fully automated 7-agent refactoring workflow. The user should not be interrupted with progress checks.

Target path: If no path is provided, use the current working directory (.).

Only ask the user a question if there is a true blocker or an explicit Stop decision.

Operating Rules

• Default: Proceed autonomously with stated assumptions
• Max interruption: One tight question only, if required to proceed safely
• Agent handoffs: Always pass full context and latest artifacts
• Gating decisions: Approve / Revise / Block (follow routing)
• Loop policy: Handle revisions automatically (max 2 cycles per gate)

---

Prerequisite Check

Before starting, verify:

git status  # Git initialized?
npm test    # Tests passing? (or equivalent)

• Git missing? Initialize it (git init, initial commit)
• Tests fail? Fix them first. Do not refactor on a broken baseline.

---

Workflow

Step 1: Explore

Invoke the @explorer agent.

Explorer produces:

• Project Understanding Document
• Architecture overview
• Current patterns and conventions
• Light notes on improvement areas

Early exit: If codebase already follows best practices → Report "No Refactoring Needed"

---

Step 2: Research

Invoke the @researcher agent with Explorer's findings.

Researcher produces:

• Best Practices Report
• Universal + project-type-specific standards
• Recommendations prioritized by impact

Early exit: If already compliant → Report "Already Compliant"

---

Step 3: Assess Tests

Invoke the @tester agent with Researcher's best practices.

Tester produces:

• Test Coverage Report
• Gaps identified
• New tests written (if needed)
• Readiness status

Stop condition: If critical areas can't be safely tested → Ask user whether to proceed with risk

---

Step 4: Plan Refactoring

Invoke the @planner agent with all upstream findings.

Planner produces:

• Refactoring Roadmap
• Phases (Small → Medium → Large)
• Specific slices with commit strategies

---

Step 4.1: Challenge the Plan (GATE)

Invoke the @challenger agent for roadmap review.

Challenger's decision:

• Approve → Proceed to Step 5
• Revise → Send to Planner, then re-run 4.1 only (max 2 cycles)
• Block → Stop. Ask user for missing info

Do NOT loop back to Steps 1-3 on "Revise."

---

Step 5: Refactor (Execution)

Invoke the @refactorer agent with approved roadmap.

Refactorer executes:

• Slice by slice
• Semantic branches per slice
• Commits per strategy
• Tests after each slice

Refactorer produces:

• Practical Summary Report
• What changed and why
• How codebase is better

---

Step 6: Verify Results (GATE)

Invoke the @verifier agent with completed work.

Verifier's decision:

• Approve → Proceed to Step 7
• Route back → Send to appropriate agent (max 2 cycles)
• Block → Stop. Escalate to user

Do NOT full re-refactor on "Route back." Fixes are targeted.

---

Step 7: Finalize + Report

Produce one concise final summary:

# Refactoring Complete: [Project Name]

## Outcome
[Shipped / Blocked]

## What Changed
- Key folders reorganized
- Key functions renamed
- Key modules extracted

## Semantic Clarity Improvements
- Naming clarity: X% → Y%
- Documentation coverage: X% → Y%
- Organization: [description]

## Tests Status
- All tests passing: ✅
- Coverage: X%

## Git History
- Total commits: N
- All merged and clean: ✅

## Follow-ups (If Any)
- [Only if truly needed]

---

Target Codebase

$ARGUMENTS