Windows POSIX shells 可能會讓 Python、PowerShell 和 cmd.exe 的中文輸出變成亂碼。此技能提供針對 Claude、Codex 及相關代理程式的 UTF-8 與 GBK 修正方式。

Weak tests can hide production bugs while inflating coverage metrics. This skill reviews test quality and identifies misleading, redundant, or excessively mocked tests.

Ambiguous specifications create implementation gaps, security risks, and costly rework. This skill reviews plans through architecture, security, quality, user, and governance perspectives before engineers begin implementation.

Complex features fail when requirements, architecture effects, and verification criteria remain unclear. This skill guides agents through research and produces an implementation-ready Gjalla specification.

Code changes can hide correctness, security, and maintainability issues before merge. This skill reviews diffs from relevant perspectives and returns prioritized, location-specific fixes.

Large feature specifications can be difficult to sequence and track. This skill creates dependency-ordered waves of small tasks with files and acceptance checks.

當結構、命名與職責逐漸偏移時，大型程式碼庫會變得難以修改。此技能會引導安全的重構步驟，在改善可維護性的同時保留既有行為。

團隊在啟動功能時，常常只有模糊的目標且缺少需求。此技能會將產品想法轉化為結構化的 PRD，並包含可衡量的標準與技術脈絡。

Spring Boot 專案可能在分層、設定、測試與安全性方面變得不一致。此技能提供精簡的最佳實務指引，協助使用 Spring Boot 設計可維護的 Java 服務。

GitHub 團隊需要針對儲存庫、議題、拉取請求、Actions 和發布版本的快速命令指引。此技能為 Claude、Codex 和 Claude Code 提供聚焦的 gh CLI 範例與工作流程。

複雜的流程與系統很難單靠文字說明。此技能可將清楚的描述轉換成可編輯的 Excalidraw 圖表及輔助摘要。

團隊經常需要符合使用者意圖的文件，涵蓋教學、操作指南、參考頁面和解釋。此技能會引導 Claude、Codex 和 Claude Code 釐清範圍、提出大綱，並起草結構化 Markdown。

控制器視窗可能會在 worker 執行期間過早關閉，或浪費上下文。此技能會安排低頻率、具 ETA 感知的檢查，並等待明確的終端狀態。

AI 輔助的分支經常包含多餘註解、不必要的防禦性程式碼，以及薄弱的型別權宜處理。此技能會檢閱分支差異，並移除不符合周圍程式碼風格的變更。

不安全或缺漏的 Claude Code 權限可能會拖慢儲存庫設定，並造成可避免的風險。此技能會檢查專案工具鏈，並提出符合偵測到的檔案與服務的設定建議。

Important project context is easily lost between sessions. This skill stores it in structured files and helps retrieve it later with PARA and qmd.

Creating a new agent requires consistent configuration, a clear role charter, and the correct approval process. This skill guides Paperclip users through discovery, drafting, review, submission, and approval follow-up.

Plans often fail when tasks lack owners, dependencies, or clear starting conditions. This skill turns accepted Paperclip plans into a verified task graph that supports focused and parallel execution.

Managing an AI-agent company can require many API calls and scattered operational checks. This skill turns board requests into structured Paperclip workflows for onboarding, approvals, monitoring, and review.

Teams need a consistent way to coordinate agent work, approvals, and task progress. This skill guides Paperclip API workflows for assignments, updates, routines, and deliverables.