AgentSkillsCN

audit-codebase

支持以可配置的角色身份开展代码库审计，生成带有严重程度评级与改进建议的 AUDIT_*.md 报告。当您需要评估项目的代码质量、安全性、性能或架构时，可调用此技能。

SKILL.md
--- frontmatter
name: audit-codebase
description: Conducts codebase audits with configurable personas, generates AUDIT_*.md reports with severity ratings and recommendations. Use when assessing code quality, security, performance, or architecture of a project.
metadata:
  author: Cahya Ong
  version: 1.0
  display-name: Audit Codebase
  keywords: [audit, codebase, quality, assessment, security, performance, architecture]
  example-prompt:
    - Audit the Restaurant codebase for code quality issues
    - Can you perform a security audit on src/Restaurant.API/?
    - Run a performance analysis of the OrderService

SKILL: Audit Codebase

Last Updated: February 5, 2026

Table of Contents

1. Overview

Conducts codebase audits using expert personas from RULE_Persona.md. Generates AUDIT_*.md reports in .ai-context/. For severity levels, see RULE_Markdown.md Section 6.

Use when: Assessing code quality, security, performance, or architecture.

Do NOT use when: Quick code review without formal report, or non-code documentation audit.

2. Audit Types

CategoryAudit Types
General SoftwareCode Quality, System Architecture, Performance, Security, Asset Management, DevOps Infrastructure, User Experience
Game DevelopmentGame Design, Level Design, Narrative Design, Audio Design, QA Testing
DocumentationTechnical Writing, Information Architecture
ResearchML Research, Data Science

See RULE_Persona.md Section 2 for persona definitions.

3. Execution Workflow

Step 1: Select Audit Type

  1. •If not specified, ask user to select from Section 2
  2. •Confirm persona perspective before proceeding
  3. •Define scope (full codebase or specific directories)

Step 2: Analyze Codebase

  1. •Recursively scan Source/ directory
  2. •High-level analysis: architectural patterns, dependencies, trends
  3. •Deep-dive analysis: specific issues with file paths, line numbers, code
  4. •Classify findings by severity per RULE_Markdown.md Section 6

Step 3: Generate Report

  1. •Check for existing AUDIT_{audit_type}.md, move to Appendix if exists
  2. •Create new report using structure (Section 4)
  3. •Include executive summary with grade and key issues

Step 4: Validate and Save

  1. •Run self-assessment protocol (Section 5)
  2. •Validate with format-markdown skill
  3. •Save to .ai-context/AUDIT_{audit_type}.md

Output (File): .ai-context/AUDIT_{audit_type}.md (template: none)

4. Report Structure

SectionContents
HeaderTitle (# {audit_type} Audit Report), Project, Date, Reviewer, Scope
Executive Summary2-3 sentence assessment, Grade (A-F), Strengths (✅), Issues (🚨❌)
High-Level AnalysisArchitectural Patterns, System-Wide Trends
Detailed AnalysisNumbered categories; each issue: path, severity, code, fix
Priority RecommendationsImmediate (Sprint), Short-term (2 Weeks), Medium-term (1 Month)
Production ReadinessStatus, Required Work estimate, Risk Areas
AppendixHistorical Audits (if applicable)

5. Self-Assessment Protocol

🚨 CRITICAL: Before presenting results, verify:

  • • Executive summary includes overall grade and key issues
  • • High-level analysis covers architectural patterns and trends
  • • Detailed analysis includes specific file paths and line numbers
  • • All severity ratings assigned consistently per RULE_Markdown.md Section 6
  • • Priority recommendations organized by timeframe
  • • Production readiness assessment provided
  • • Historical audits moved to Appendix if applicable
  • • Each recommendation has specific action description
  • • Code examples show both problem and solution
  • • Run format-markdown skill on generated file
  • • Re-read generated file to confirm accuracy

6. Historical Tracking

Archive Format:

  • •Section title: ## Appendix: Historical Audits
  • •Subsection: ### Audit from [Month DD, YYYY]
  • •Order: Latest to oldest

Guidelines: Track resolved issues and grade progression, note newly introduced problems, compare metrics over time.