skills

Overview of skills

What it is

Skills is a collection of agent skills for Claude Code. Each skill is a folder containing a SKILL.md with frontmatter (name, description) and instructions that Claude loads on demand when the task matches. The repository also includes an AGENTS.md workflow template.

Key features

• before-and-after: Captures before/after screenshots of web pages or elements and outputs a PR-ready markdown comparison table using the @vercel/before-and-after CLI.- code-structure: Provides service layer architecture guidance, preferring actions for orchestration and services for reusable operations when no established convention exists.- evidence-driven-testing: Records visual proof while testing UI behavior using a bundled recorder (scripts/evidence.py) that captures sessions and posts video and results summary to the PR and tracker issue.- greploop: Addresses Greptile feedback on a PR, MR, or shelved changelist, targeting a fresh 5/5 review within a configurable number of iterations (default 10).- greploop-apps: A compatibility entrypoint for greploop --trigger @greptile-apps, using the same workflow and tested GitHub helper.- new-feature: Isolates repository edits in a Git worktree based on origin/main, allowing multiple agents to work the same repository concurrently.- unslop: Edits prose to remove AI tells and put a human voice back in, catching 31 patterns (puffery, filler, hedging, chatbot phrases, etc.) and applying a four-step loop.

Use cases

• Starting a feature, fix, or documentation edit while isolating changes in a Git worktree.
• Capturing visual proof of UI changes for PRs with before/after comparison tables.
• Deciding what belongs in actions vs. shared services when multiple workflows duplicate operational logic.
• Recording visual evidence while testing UI behavior via computer use or scripted captures.
• Getting a PR to a clean Greptile review before merge.
• Editing prose (commit messages, PR titles, docs, code comments) to remove AI tells and add human voice.

Getting started

bashnpx skills add michaelshimeles/skillsClaude Code picks up the skill automatically and invokes it when a task matches the skill's description. You can also invoke one explicitly with /code-structure or /evidence-driven-testing.The alternate Greptile entrypoint requires both folders:bashcp -r greploop greploop-apps ~/.claude/skills/

License and pricing

The license is not stated for the overall repository. Some vendored skills include their own licenses: before-and-after (PolyForm Shield 1.0.0), greploop and greploop-apps (MIT), unslop (MIT). The pricing type is open source.

這是什麼

Skills 是一組為 Claude Code 設計的代理技能集合。每個技能是一個包含 SKILL.md 的資料夾，內有 frontmatter（名稱、描述）和指令，當任務匹配時 Claude 會按需載入。此儲存庫也包含一個 AGENTS.md 工作流程範本。

主要功能

• before-and-after：擷取網頁或元素的前後對比截圖，並使用 @vercel/before-and-after CLI 輸出 PR 就緒的 Markdown 比較表格。- code-structure：提供服務層架構指引，在專案沒有既定慣例時偏好使用 actions 進行編排，並使用 services 處理可重複的操作。- evidence-driven-testing：在測試 UI 行為時記錄視覺證據，使用內建的錄製器（scripts/evidence.py）擷取工作階段，並將影片和結果摘要發佈到 PR 和追蹤議題。- greploop：處理 PR、MR 或擱置變更清單上的 Greptile 回饋，目標是在可設定的迭代次數內（預設 10 次）獲得當前修訂版本的 5/5 審查。- greploop-apps：greploop --trigger @greptile-apps 的相容性進入點，使用相同的工作流程和經過測試的 GitHub 輔助工具。- new-feature：基於 origin/main 在 Git worktree 中隔離儲存庫編輯，允許多個代理同時處理同一個儲存庫。- unslop：編輯散文以移除 AI 痕跡並恢復人類語氣，捕捉 31 種模式（浮誇、填充、迴避、聊天機器人用語等），並應用四步驟循環。

使用情境

• 開始一個功能、修復或文件編輯，同時在 Git worktree 中隔離變更。
• 為 PR 擷取 UI 變更的視覺證明，產生前後對比表格。
• 當多個工作流程重複相同的操作邏輯時，決定哪些屬於 actions 哪些屬於共享 services。
• 透過電腦使用或腳本擷取來記錄測試 UI 行為時的視覺證據。
• 在合併前讓 PR 獲得乾淨的 Greptile 審查。
• 編輯散文（提交訊息、PR 標題、文件、程式碼註解）以移除 AI 痕跡並加入人類語氣。

快速開始

bashnpx skills add michaelshimeles/skillsClaude Code 會自動選取技能，並在任務符合技能描述時呼叫它。您也可以使用 /code-structure 或 /evidence-driven-testing 明確呼叫。替代的 Greptile 進入點需要兩個資料夾：bashcp -r greploop greploop-apps ~/.claude/skills/

授權與價格

整體儲存庫的授權未說明。部分外購技能包含其自有授權：before-and-after（PolyForm Shield 1.0.0）、greploop 和 greploop-apps（MIT）、unslop（MIT）。定價類型為開源。

Based on the project's public README and repository metadata.