Fix

---
description: "Agentic auto-fix - Parallel scan with autonomous correction"
argument-hint: "[--dry] [--sequential] [--level N] [file_path] | --resume [ID]"
type: utility
allowed-tools: Task, AskUserQuestion, TodoWrite, Bash, Read, Write, Edit, Glob, Grep
model: inherit
---

## Pre-execution Context

!git status --porcelain
!git diff --name-only HEAD

## Essential Files

@.moai/config/sections/ralph.yaml

---

# /moai:fix - Agentic Auto-Fix

## Core Principle: Fully Autonomous Fixing

AI autonomously finds and fixes issues.

START: Issue Detection ↓ AI: Parallel Scan → Classify → Fix → Verify ↓ AI: Add Completion Marker

## Command Purpose

Autonomously fix LSP errors and linting issues:

1. **Parallel Scan** (LSP + AST-grep + Linters simultaneously)
2. **Auto Classification** (Level 1-4)
3. **Auto Fix** (Level 1-2)
4. **Verification**
5. **Report**

Target: $ARGUMENTS

## Quick Start

```bash
# Default fix (parallel scan)
/moai:fix

# Sequential scan (for debugging)
/moai:fix --sequential

# Preview only
/moai:fix --dry

# Specific file
/moai:fix src/auth.py

# Limit fix level
/moai:fix --level 2

# Resume from last snapshot
/moai:fix --resume

# Resume from specific snapshot
/moai:fix --resume fix-20260119-143052

Command Options

Option	Alias	Description	Default
--dry	--dry-run	Preview only	Apply
--sequential	--seq	Sequential scan (for debugging)	Parallel
--level N	-	Maximum fix level	3
--errors	--errors-only	Fix errors only	All
--security	--include-security	Include security issues	Exclude
--no-fmt	--no-format	Skip formatting	Include
--resume [ID]	--resume-from	Resume from snapshot	Latest

Parallel Scan

# Sequential (30s)
LSP → AST → Linter

# Parallel (8s)
LSP   ├─┐
      ├─→ Merge (3.75x faster)
AST   ├─┤
     ├─┘
Linter

Parallel Scan Implementation

By default, execute all diagnostic tools simultaneously for optimal performance:

Step 1 - Launch Background Tasks:

1. LSP Diagnostics: Use Bash tool with run_in_background set to true for LSP diagnostic command based on detected language
2. AST-grep Scan: Use Bash tool with run_in_background set to true for ast-grep with security rules from sgconfig.yml
3. Linter Scan: Use Bash tool with run_in_background set to true for appropriate linter (ruff, eslint, etc.)

Step 2 - Collect Results:

1. Use TaskOutput tool to collect results from all three background tasks
2. Wait for all tasks to complete (timeout: 120 seconds per task)
3. Handle partial failures gracefully - continue with available results

Step 3 - Aggregate and Deduplicate:

1. Parse output from each tool into structured issue list
2. Remove duplicate issues appearing in multiple scanners
3. Sort by severity: Critical, High, Medium, Low
4. Group by file path for efficient fixing

Step 4 - Proceed to Classification:

1. Classify aggregated issues into Levels 1-4
2. Call TodoWrite with all issues as pending items
3. Begin fix execution

Language-Specific Commands:

Python: ruff check --output-format json for linter, mypy --output json for types TypeScript: eslint --format json for linter, tsc --noEmit for types Go: golangci-lint run --out-format json for linter Rust: cargo clippy --message-format json for linter

Auto-Fix Levels

Level	Description	Approval	Examples
1	Immediate	Not required	import, whitespace
2	Safe	Log only	rename var, add type
3	Review	Required	logic, API
4	Manual	Not allowed	security, architecture

TODO-Obsessive Rule

[HARD] TodoWrite Tool Mandatory Usage:

1. Immediate Creation: When issues are discovered, call TodoWrite tool to add items with pending status
2. Immediate Progress: Before starting work, call TodoWrite tool to change item to in_progress
3. Immediate Completion: After completing work, call TodoWrite tool to change item to completed
4. Prohibited: Output TODO lists as text (MUST use TodoWrite tool)

WHY: Using TodoWrite tool allows users to track progress in real-time.

Output Format

Preview

## Fix: Dry Run

### Scan (0.8s, parallel)
- LSP: 12 issues
- AST-grep: 0 security
- Linter: 5 issues

### Level 1 (12 items)
- src/auth.py: import, formatting
- src/api/routes.py: import order
- tests/test_auth.py: whitespace

### Level 2 (3 items)
- src/auth.py:45 - 'usr' → 'user'
- src/api/routes.py:78 - add type
- src/models.py:23 - dataclass?

### Level 4 (2 items)
- src/auth.py:67 - logic error
- src/api/routes.py:112 - SQL injection

No changes (--dry).

Complete

## Fix: Complete

### Applied
- Level 1: 12 issues
- Level 2: 3 issues
- Level 3: 0 issues

### Evidence
**src/auth.py:5** - Removed unused `os`, `sys`
**src/auth.py:23** - Fixed whitespace
**src/api/routes.py:12** - Sorted imports

### Remaining (Level 4)
1. src/auth.py:67 - logic error
2. src/api/routes.py:112 - SQL injection

### Next
/moai:loop  # Continue with loop

Quick Reference

# Fix (default parallel)
/moai:fix

# Sequential scan
/moai:fix --sequential

# Preview only
/moai:fix --dry

# Errors only
/moai:fix --errors

# Specific file
/moai:fix src/auth.py

---

EXECUTION DIRECTIVE

1. Parse $ARGUMENTS (extract --sequential, --dry, --level, --errors, --security flags)

2. Detect project language from indicator files (pyproject.toml, package.json, go.mod, Cargo.toml)

3. Execute diagnostic scan:

   IF --sequential flag is specified:

   3a. Run LSP diagnostics, then AST-grep, then Linter sequentially

   ELSE (default parallel mode):

   3b. Launch all three diagnostic tools in parallel using Bash with run_in_background: - Task 1: LSP diagnostics for detected language - Task 2: AST-grep scan with sgconfig.yml rules - Task 3: Linter for detected language 3c. Collect results using TaskOutput for each background task

   3