with one click
skills
skills contains 10 collected skills from bennytsai1234, with repository-level occupation coverage and site-owned skill detail pages.
Skills in this repository
Initialize or rebuild a repository atlas under docs/ for AI-assisted code navigation. Only for an explicit atlas build, rebuild, refresh, or rescan requested by a human — never for ordinary development, and never inside a delegated subagent task.
把 YouTube 影片網址轉成帶時間戳的繁體中文逐字稿。當使用者貼出影片連結,並想要字幕、逐字稿、影片文案、把影片內容轉成文字、或把影片內容載入對話中討論時,務必使用此 skill。流程為:優先抓取影片現成字幕,抓不到才用本地語音辨識(MOSS + 靜音切片),最後統一簡轉繁體並輸出帶時間戳的文字檔,再把全文帶進對話。
Update or diagnose the OpenAI Codex CLI on this controlled Windows environment. Use for version checks, CLI updates, and update failures. Never bypass TLS or change global certificate settings.
Diagnose and configure Windows Chinese/Japanese/Korean font fallback, especially replacing Noto Sans SC/TC and Noto Sans CJK SC/TC with Sarasa/Gengsha UI fonts system-wide. Use when users report mixed simplified/traditional Chinese font shapes in VS Code, Electron/Chromium apps, Markdown/extension README pages, or other Windows apps; when Noto Sans SC/TC must be substituted; or when ClearType color fringing appears after CJK font changes.
對任一程式碼庫執行全面源碼安全稽核(A–E:相依漏洞、SAST、機密外洩、人工稽核),並提供修補劇本。針對銀行管控 Windows 環境調校。當使用者要求源碼安全檢查、安全稽核、漏洞掃描、機密外洩檢查時使用。
從一個開發順利的成功專案中,透過訪談使用者 + 掃描程式碼,交叉萃取出「讓 AI agent 開發很順」的可遷移、語言/框架無關的個人開發準則,產出可直接放進 ~/.claude/CLAUDE.md 的全域規範,以及一份維護指南。當使用者想把既有專案的成功經驗沉澱成跨專案可複用的全域 AI 協作規範時啟動。觸發語句包含:「整理成開發規範」「提取可復用的準則」「生成 CLAUDE.md」「全域 CLAUDE.md」「把這個專案蒸餾成 skill / 規範」「讓未來的專案也能用這套做法」「探索專案整理成 agent 規範」。即使使用者只說「把這個專案整理成規範」而沒明講「全域」或「可遷移」,只要意圖是沉澱跨專案經驗,也應啟動。
Use when the user asks to install / add a Claude Code skill ("install this skill", "add this skill", "安裝這個 skill", "幫我裝 skill"). Installs by SYMLINKING the skill folder from /home/benny/skills (the canonical git source of truth) into ~/.claude/skills/<name> — never by copying — so `git pull` updates propagate automatically. Handles the varying source layouts and the /plugin exception for marketplace-packaged collections.
Use when operating Hermes Agent itself: checking the version, updating from upstream, running doctor/doctor --fix, installing the browser stack, setting tokens, managing the gateway service, and telling apart real fixables from by-design 'optional integration not configured' doctor noise.
Use when operating OpenClaw itself: checking versions, updating the CLI, managing the gateway service, verifying health, handling restart edge cases after upgrades, cleaning up plugins/stale install-index state, and migrating plaintext secrets to SecretRefs.
Use when the user wants to plan a software product (web app, SaaS dashboard, mobile app, marketing site, content site, API service, browser extension, or desktop app) before any code is written. Produces SDLC Stage 1 (requirements) and Stage 2 (iterative visual prototype) artifacts plus a handoff package for downstream coding agents like Claude Code or Codex. Use when the user has a vague or concrete product idea and wants the final implementation to match their mental image with high fidelity.