スキル一覧に戻る

review-codebase

pjt222
更新日 2 days ago
4 閲覧
17
2
17
GitHubで表示
メタdesign

について

このスキルは、コードベース全体に対して包括的かつ多段階のレビューを実行し、アーキテクチャ、セキュリティ、コード品質、UX/アクセシビリティを単一の調整されたパスで分析します。優先順位付けされた発見事項と重大度評価を構造化された表として出力し、GitHubのイシューに直接変換できる形式で提供します。個別の変更や単一ドメインのレビューではなく、深く全体的な監査のためにご利用ください。

クイックインストール

Claude Code

推奨
メイン
npx skills add pjt222/agent-almanac -a claude-code
プラグインコマンド代替
/plugin add https://github.com/pjt222/agent-almanac
Git クローン代替
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/review-codebase

このコマンドをClaude Codeにコピー&ペーストしてスキルをインストールします

ドキュメント

Review Codebase

Multi-phase deep codebase review producing severity-rated findings with fix-order recommendations. Unlike review-pull-request (scoped to a diff) or single-domain reviews (security-audit-codebase, review-software-architecture), this skill covers entire project or subproject across all quality dimensions in one pass.

When Use

  • Whole-project or subproject review (not PR-scoped)
  • New codebase onboarding — building mental model of what exists and what needs attention
  • Periodic health checks after sustained development
  • Pre-release quality gate across architecture, security, code quality, UX
  • When output should feed direct into issue creation or sprint planning

Inputs

  • Required: target_path — root directory of codebase or subproject to review
  • Optional:
    • scope — which phases to run: full (default), security, architecture, quality, ux
    • output_formatfindings (table only), report (narrative), both (default)
    • severity_threshold — minimum severity to include: LOW (default), MEDIUM, HIGH, CRITICAL

Steps

Step 1: Census

Inventory codebase to establish scope and identify review targets.

  1. Count files by language/type: find target_path -type f | sort by extension
  2. Measure total line counts per language
  3. ID test directories and estimate test coverage (files with tests vs files without)
  4. Check dependency state: lockfiles present, outdated dependencies, known vulnerabilities
  5. Note build system, CI/CD configuration, documentation state
  6. Record census as opening section of report

Got: Factual inventory — file counts, languages, test presence, dependency health. No judgments yet.

If fail: Target path empty or inaccessible? Stop and report. Specific subdirectories inaccessible? Note them and continue with what is available.

Step 2: Architecture Review

Assess structural health: coupling, duplication, data flow, separation of concerns.

  1. Map module/directory structure. ID primary architectural pattern
  2. Check for code duplication — repeated logic across files, copy-paste patterns
  3. Assess coupling — how many files must change for single feature modification
  4. Evaluate data flow — clear boundaries between layers (UI, logic, data)?
  5. ID dead code, unused exports, orphaned files
  6. Check for consistent patterns — does codebase follow its own conventions?
  7. Rate each finding: CRITICAL, HIGH, MEDIUM, or LOW

Got: List of architectural findings with severity ratings and file references. Common findings: mode dispatch duplication, missing abstraction layers, circular dependencies.

If fail: Codebase too small for meaningful architecture review (< 5 files)? Note this and skip to Step 3. Architecture review needs enough code to have structure.

Step 3: Security Audit

Identify security vulnerabilities and defensive coding gaps.

  1. Scan for injection vectors: HTML injection (innerHTML), SQL injection, command injection
  2. Check authentication and authorization patterns (if applicable)
  3. Review error handling — errors silently swallowed? Error messages leak internals?
  4. Audit dependency versions against known CVEs
  5. Check for hardcoded secrets, API keys, credentials
  6. Review Docker/container security: root user, exposed ports, build secrets
  7. Check localStorage/sessionStorage for sensitive data storage
  8. Rate each finding: CRITICAL, HIGH, MEDIUM, or LOW

Got: List of security findings with severity, affected files, remediation guidance. CRITICAL findings include injection vulnerabilities and exposed secrets.

If fail: No security-relevant code exists (pure documentation project)? Note this and skip to Step 4.

Step 4: Code Quality

Evaluate maintainability, readability, defensive coding.

  1. ID magic numbers and hardcoded values that should be named constants
  2. Check for consistent naming conventions across codebase
  3. Find missing input validation at system boundaries
  4. Assess error handling patterns — consistent? Provide useful messages?
  5. Check for commented-out code, TODO/FIXME markers, incomplete implementations
  6. Review test quality — tests testing behavior or implementation details?
  7. Rate each finding: CRITICAL, HIGH, MEDIUM, or LOW

Got: List of quality findings focused on maintainability. Common findings: magic numbers, inconsistent patterns, missing guards.

If fail: Codebase generated or minified? Note this and adjust expectations. Generated code has different quality criteria than hand-written code.

Step 5: UX and Accessibility (if frontend exists)

Evaluate user experience and accessibility compliance.

  1. Check ARIA roles, labels, landmarks on interactive elements
  2. Verify keyboard navigation — can all interactive elements be reached via Tab?
  3. Test focus management — does focus move logical when panels open/close?
  4. Check responsive design — test at common breakpoints (320px, 768px, 1024px)
  5. Verify color contrast ratios meet WCAG 2.1 AA standards
  6. Check screen reader compatibility — dynamic content changes announced?
  7. Rate each finding: CRITICAL, HIGH, MEDIUM, or LOW

Got: List of UX/a11y findings with WCAG references where applicable. No frontend exists? This step produces "N/A — no frontend code detected."

If fail: Frontend code exists but cannot be rendered (missing build step)? Audit source code statically and note that runtime testing was not possible.

Step 6: Findings Synthesis

Compile all findings into prioritized summary.

  1. Merge findings from all phases into single table
  2. Sort by severity (CRITICAL first, then HIGH, MEDIUM, LOW)
  3. Within each severity level, group by theme (security, architecture, quality, UX)
  4. For each finding, include: severity, phase, file(s), one-line description, suggested fix
  5. Produce recommended fix order that considers dependencies between fixes
  6. Summarize: total findings by severity, top 3 priorities, estimated effort level

Got: Findings table with columns: #, Severity, Phase, File(s), Finding, Fix. Fix-order recommendation that accounts for dependencies (e.g., "refactor architecture before adding tests").

If fail: No findings produced? This is itself a finding — either codebase exceptionally clean or review too shallow. Re-examine at least one phase with deeper inspection.

Checks

  • All requested phases completed (or explicit skipped with justification)
  • Every finding has severity rating (CRITICAL/HIGH/MEDIUM/LOW)
  • Every finding references at least one file or directory
  • Findings table sorted by severity
  • Fix-order recommendations account for dependencies between findings
  • Summary includes total counts by severity
  • If output_format includes report, narrative sections accompany table

Scaling with Rest

Between review phases, use /rest as checkpoint — especially between phases 2-5, which need different analytical perspectives. Checkpoint rest (brief, transitional) prevents momentum of one phase from biasing next. See rest skill "Scaling Rest" section for guidance on checkpoint vs full rest.

Pitfalls

  • Boil the ocean: Review every line of large codebase produces noise. Focus on high-impact areas: entry points, security boundaries, architectural seams
  • Severity inflation: Not every finding is CRITICAL. Reserve CRITICAL for exploitable vulnerabilities and data-loss risks. Most architectural issues are MEDIUM
  • Miss the forest for the trees: Individual code quality issues matter less than systemic patterns. Magic numbers appear in 20 files? That is one architectural finding, not 20 quality findings
  • Skip the census: Census (Step 1) seems bureaucratic but prevents reviewing code that does not exist or missing entire directories
  • Phase bleed: Security findings during architecture review, or quality findings during security audit. Note them for correct phase rather than mix concerns — produces cleaner findings table

See Also

  • security-audit-codebase — deep-dive security audit when review-codebase security phase reveals complex vulnerabilities
  • review-software-architecture — detailed architecture review for specific subsystems
  • review-ux-ui — comprehensive UX/accessibility audit beyond what phase 5 covers
  • review-pull-request — diff-scoped review for individual changes
  • clean-codebase — implements code quality fixes identified by this review
  • create-github-issues — converts findings table into tracked GitHub issues

GitHub リポジトリ

pjt222/agent-almanac
パス: i18n/caveman/skills/review-codebase
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

関連スキル

content-collections

メタ

このスキルは、Content Collections(Markdown/MDXファイルを型安全なデータコレクションに変換するTypeScriptファーストのツール)の本番環境でテストされた設定を提供します。Zodバリデーションによる型安全性を実現し、ブログ、ドキュメントサイト、コンテンツ重視のVite + Reactアプリケーション構築時にご利用ください。Viteプラグインの設定、MDXコンパイルから、デプロイ最適化、スキーマバリデーションまで、すべてを網羅しています。

スキルを見る

polymarket

メタ

このスキルは、開発者がPolymarket予測市場プラットフォームを活用したアプリケーション構築を可能にします。API統合による取引や市場データの取得に加え、WebSocketを介したリアルタイムデータストリーミングにより、ライブ取引や市場活動を監視できます。取引戦略の実装や、ライブ市場更新を処理するツールの作成にご利用ください。

スキルを見る

creating-opencode-plugins

メタ

このスキルは、開発者がコマンド、ファイル、LSP操作など25種類以上のイベントタイプにフックするOpenCodeプラグインを作成することを支援します。JavaScript/TypeScriptモジュール向けに、プラグイン構造、イベントAPI仕様、および実装パターンを提供します。カスタムイベント駆動ロジックでOpenCode AIアシスタントのライフサイクルをインターセプト、監視、または拡張する必要がある場合にご利用ください。

スキルを見る

sglang

メタ

SGLangは、高性能なLLMサービングフレームワークであり、RadixAttentionプレフィックスキャッシュを活用したJSON、正規表現、エージェントワークフロー向けの高速で構造化された生成を特長とします。特にプレフィックスが繰り返されるタスクにおいて、大幅に高速な推論を実現し、複雑な構造化出力やマルチターン対話に最適です。制約付きデコードが必要な場合や、広範なプレフィックス共有を伴うアプリケーションを構築する場合は、vLLMなどの代替案ではなくSGLangを選択してください。

スキルを見る