code-quality-audit（代码质量审计）

Overview

This skill provides two-phase code quality review:

1. Audit Phase (Sukuna mindset): Cold, systematic dissection of code quality, complexity, duplication, and performance anti-patterns. Identifies fatal flaws with surgical precision.
2. Improvements Phase (Gojo mindset): Elegant, comprehensive refactoring roadmap organized by priority and impact. Offers perceptive solutions.

Both outputs are produced in a single run, enabling teams to understand what's broken in their implementation and how to fix it systematically.

Activation Rules

Explicit triggers only. This skill activates ONLY when the user explicitly mentions:

• "audit code"
• "code quality audit"
• "brutal code review"
• "Sukuna code review"
• Similar explicit phrases requesting a code quality review.

Required input: User must provide or reference source code files (e.g., Python, JS, Go, etc.) or a directory. If no files are provided, respond normally and politely request: "Please provide the source code files you'd like audited."

File format: All audited content is source code. Do NOT attempt to run static analysis tools if not available, rely on your deep understanding of code principles.Keep both files in code-audit folder.

Phase 1: Audit Report (audit.md)

Mindset: Ruthless, Direct, Factual

The audit phase adopts a no-nonsense, penetrating analytical approach:

• Direct language: Identify exactly what is wrong and why it matters.
• Systematic: Analyze complexity, style adherence, readability, and performance within and across files.
• Evidence-based: Every claim is backed by specific file names, line numbers, or code snippets.
• No sugar-coating: Call out spaghetti code plainly, but always with actionable reasoning.

Output Structure

Generate audit.md with this exact structure:

# Code Quality Audit Report

## Executive Summary
- **Overall Score**: X/1000
- **Maintainability Verdict**: [Maintainable / Requires Refactoring / Unmaintainable (Rewrite Suggested)]
- **Primary Strengths**: ...
- **Critical Weaknesses**: ...

## File/Component Scores
| File/Path | Score /100 | Assessment |
|-----------|------------|------------|
| [file] | [score] | [one-line verdict] |

## Detailed Findings

### Complexity & Duplication
[Findings on cognitive complexity, overly nested loops, long methods, DRY violations]

### Style & Convention Adherence
[Analysis of naming conventions, consistency, idiomatic usage (e.g., Pythonic code)]

### Readability & Maintainability
[Assessment of self-documenting code vs over-reliance on comments, clarity of logic]

### Performance Anti-patterns
[O(N^2) loops where O(N) is possible, memory leaks, inefficient data structures]

### Security & Error Handling
[Swallowed exceptions, hardcoded secrets, injection vectors, poor validation]

## Final Verdict
[Summary of overall health and whether major refactoring is needed]

Scoring Methodology

Per-file scores (/100):

• 90–100: Pristine, idiomatic execution, production-ready
• 70–89: Solid implementation with exploitable flaws or minor tech debt
• 50–69: Functional but messy; high complexity or duplication
• <50: Dangerous; unmaintainable, actively harmful

Overall score (/1000):

• Weighted average of file scores, with severe penalties for systemic anti-patterns (-100 to -400)
• Rarely exceeds 700–800 unless code is near-excellent
• Calculation: (Sum of weighted file scores) - (Penalties for systemic issues)

Evaluation criteria:

• Cognitive and cyclomatic complexity
• Don't Repeat Yourself (DRY) and Keep It Simple (KISS) adherence
• Clean, consistent naming conventions
• Safe and robust error handling
• Idiomatic language usage
• Proper separation of business logic from I/O or frameworks

Standards for Analysis

• Reference everything: Use exact line numbers and code snippets to ground every claim.
• Architect-level thinking applied to implementation: Don't just point out a missing docstring; point out if a 500-line class should be split into smaller, cohesive modules.
• Balance rigor with usefulness: Always explain the *impact* of an issue (e.g., "This massive if/else block makes unit testing impossible and violates Open/Closed principle").
• Do not hallucinate: Only discuss what is actually present in the provided source code.
• Report everything you observe: If code is fragile, poorly structured, or dangerous, include it regardless of how standard the pattern appears. Do not rationalize away findings as "pragmatic" or "good enough." If you noticed it, report it.
• Output goes under the user's name: Your audit reflects on them professionally. Score honestly — inflated scores help no one.

Phase 2: Improvements Roadmap (improvements.md)

Mindset: Perceptive, Solution-Focused, Comprehensive

The improvements phase adopts a calm, methodical, strategic mindset:

• Clarity of vision: See the entire codebase landscape and prioritize refactors precisely.
• Elegance: Offer refined, idiomatic, and robust refactoring solutions.
• Comprehensiveness: Address blocking tech debt first, then medium-term improvements.
• Actionability: Provide clear direction for implementation.

Output Structure

Generate improvements.md with this exact structure:

# Refactoring Improvements Roadmap

## Critical Refactors
[Issues that must be fixed immediately; they block extensibility, hurt performance, or cause bugs]

### Refactor: [Name]
- **Location**: [File/line reference]
- **Problem**: [Clear description]
- **Impact**: [Why this is critical]
- **Suggested Approach**: [Solution outline, with short code snippet if useful]

## Medium Priority Improvements
[Issues that degrade quality or maintainability over time]

### Refactor: [Name]
- **Location**: [File/line reference]
- **Problem**: [Clear description]
- **Impact**: [Why this matters]
- **Suggested Approach**: [Solution outline/Snippet]

## Nice-to-Have Enhancements
[Modernization, type-hinting improvements, or minor style polishes]

### Enhancement: [Name]
- **Location**: [File/line reference]
- **Description**: [What could be improved]
- **Benefit**: [Why it's worth doing]
- **Suggested Approach**: [Solution outline]

Standards for Improvements

• Prioritization is precise: Critical refactors fix bugs or untangle major spaghetti logic. Medium issues polish logic. Nice-to-have issues modernize.
• Solutions are reasoned: Each suggestion explains why and how (e.g., "Extracting this nested loop into a helper method reduces cognitive complexity and enables direct unit testing").
• No full code generation yet: Improvements are refactoring strategies/proposals, not a complete rewrite of the files. Short snippets are okay for illustration.
• Bridge to action: Each improvement should be scoped such that a developer can understand exactly what function/class to change next.

Workflow

1. User triggers: User provides source code files and says "audit code" or "perform a code quality audit".
2. Agent reads files: Parse all provided source code.
3. Sukuna phase: Produce audit.md with ruthless, systematic code breakdown.
4. Gojo phase: Produce improvements.md with solution-focused refactoring roadmap.
5. Deliver both: Present both files, written to disk, and ask if the user wants to execute any specific refactors.

Quality Assurance

• Do not invoke this skill unless explicitly triggered.
• Keep output highly professional, devoid of JJK flavor text in the final markdown.
• Do not auto-fix the code entirely without asking—this is planning and analysis only.
• If files are too large, focus on the most complex files and call out the limitations.

Example Trigger Phrases

• "Audit this code"
• "Run a quality audit on these python files"
• "Give me a brutal code review"
• "Audit code quality for [list of files]"