review-codebase
About
This skill performs comprehensive multi-phase reviews of entire codebases or subprojects, analyzing architecture, security, code quality, and UX/accessibility in a single pass. It generates prioritized findings with severity ratings in a structured format ready for GitHub issue creation. Use it for deep codebase analysis, pre-release quality gates, or onboarding to new projects.
Quick Install
Claude Code
Recommendednpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/review-codebaseCopy and paste this command in Claude Code to install this skill
Documentation
審庫
多階深庫審生重級發見與修序薦。異 review-pull-request(限 diff)或單域審(security-audit-codebase、review-software-architecture)、此技覆全項或子項跨諸質維於一過。
用
- 全項或子項審(非 PR 限)
- 新庫熟——建心模於存何與何需顧
- 久開後周健察
- 發前質閘跨構、安、碼質、UX
- 出宜直入問建或衝刺計
入
- 必:
target_path— 欲審庫或子項根錄 - 可:
scope— 行階:full(默)、security、architecture、quality、uxoutput_format—findings(唯表)、report(敘)、both(默)severity_threshold— 含最低重:LOW(默)、MEDIUM、HIGH、CRITICAL
行
一:census
錄庫以立範識審標。
- 按語/型計檔:
find target_path -type f | sort by extension - 量各語總行
- 識測錄、估測覆(有測檔對無)
- 察依態:鎖檔存、舊依、知漏
- 記建系、CI/CD 配、文態
- census 為報首段
得:實錄——檔數、語、測在、依健。未判。
敗:標路空或不可達→止報。某子錄不可達→記之以所有續。
二:構審
估構健:耦、複、數流、關分。
- 圖模/錄構、識主構模
- 察碼複——跨檔重邏、複貼模
- 估耦——一功改需變諸檔幾
- 評數流——層間(UI、邏、數)有明界乎?
- 識死碼、未用出、孤檔
- 察恆模——庫循己約乎?
- 各發見分:CRITICAL、HIGH、MEDIUM、LOW
得:構發見列、含重級與檔引。常發見:模派複、缺抽層、環依。
敗:庫過小(< 5 檔)→記略入三。構審需碼足以有構。
三:安審
識安漏與防碼缺。
- 掃注向:HTML 注(
innerHTML)、SQL 注、命注 - 察認與授模(適用)
- 審錯處——錯默吞乎?錯訊漏內乎?
- 審依版對知 CVE
- 察硬密、API 鍵、憑
- 審 Docker/容安:根用、露埠、建密
- 察 localStorage/sessionStorage 為敏數存
- 各發見分:CRITICAL、HIGH、MEDIUM、LOW
得:安發見列、含重、影檔、修導。CRITICAL 含注漏與露密。
敗:無安相碼(純文項)→記略入四。
四:碼質
評維、易讀、防碼。
- 識魔數與宜為命常之硬值
- 察跨庫恆命約
- 尋系界缺輸驗
- 估錯處模——恆乎?供有用訊乎?
- 察註碼、TODO/FIXME 標、未全實
- 審測質——測為乎抑為實詳?
- 各發見分:CRITICAL、HIGH、MEDIUM、LOW
得:質發見列注於維。常發見:魔數、不恆模、缺護。
敗:庫為生或縮→記、調期。生碼有異於手碼之質準。
五:UX 與可達(前端存)
評用驗與可達合。
- 察互素 ARIA 角、標、地標
- 驗鍵盤導——諸互素皆可 Tab 至乎?
- 測焦管——板開閉時焦邏動乎?
- 察響應設——測常斷點(320px、768px、1024px)
- 驗色對比合 WCAG 2.1 AA 標
- 察屏讀容——動容變宣告乎?
- 各發見分:CRITICAL、HIGH、MEDIUM、LOW
得:UX/a11y 發見列、適用處引 WCAG。前端無→此步生「N/A — no frontend code detected」。
敗:前端碼存而不可渲(缺建步)→靜析源碼、記運時測不能。
六:發見合
合諸發見入序摘。
- 諸階發見合一表
- 按重排(CRITICAL 先、後 HIGH、MEDIUM、LOW)
- 各重級內、按題聚(安、構、質、UX)
- 各發見含:重、階、檔、一句述、擬修
- 生薦修序、顧修間依
- 摘:按重總計、首三優、估力級
得:發見表含列:#、Severity、Phase、File(s)、Finding、Fix。修序薦顧修間依(如「測前重構構」)。
敗:無發見生→自為發見——或庫異潔或審過淺。再深察至少一階。
驗
- 諸請階皆畢(或明略含由)
- 各發見有重級(CRITICAL/HIGH/MEDIUM/LOW)
- 各發見引至少一檔或錄
- 發見表按重排
- 修序薦顧發見間依
- 摘含按重總計
-
output_format含report→敘段伴表
縮息
審階間、/rest 為檢——尤二至五階間、需異析角。檢息(簡、轉)防一階勢偏次。見 rest 技「Scaling Rest」段為檢與全息導。
忌
- 沸海:審大庫每行生噪。注高影域:入點、安界、構接
- 重膨:非皆 CRITICAL。CRITICAL 留可剝漏與數失險。多構患為 MEDIUM
- 失林為樹:個碼質患少於系模重。魔數現於 20 檔→一構發見、非二十質發見
- 略 census:census(一)似官而防審不存碼或失整錄
- 階滲:構審中安發見、或安審中質發見。為正階記、勿混——生潔發見表
參
security-audit-codebase— 審庫安階揭複漏時深安審review-software-architecture— 特子系詳構審review-ux-ui— 過五階之全 UX/可達審review-pull-request— diff 限審為個變clean-codebase— 實此審識之碼質修create-github-issues— 化發見表為追 GitHub 問
GitHub Repository
Related Skills
content-collections
MetaThis skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.
polymarket
MetaThis skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.
creating-opencode-plugins
MetaThis skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
