Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add fox100039-design/bugezy --skill docsgit clone --depth 1 https://github.com/fox100039-design/bugezyWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/fox100039-design/bugezy/docs)<a href="https://agentmods.dev/skills/fox100039-design/bugezy/docs"><img src="https://agentmods.dev/badge/skills/fox100039-design/bugezy/docs.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00101 | $0.04296 |
| Opus 5 | $0.00051 | $0.02148 |
| Sonnet 5 | $0.00020 | $0.00859 |
| Haiku 4.5 | $0.00010 | $0.00430 |
Grade A, and why
BugEzy MCP Voice Debug Tool Dev Guide scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- AI 語音:精簡 `/api/summarize`(Llama 3.3,轉重點)+ 校正 `/api/correct`(Llama 3.3,修錯字/去贅字/還原術語、保留原意,PM-60c)。踩雷:測中文 API 用 Python 送 UTF-8,別用 Windows Git-Bash curl(會把 body 編碼壞 + 終端機顯示花掉) How it starts
The opening of the file, as written. The whole thing — 146 lines — stays where its author put it; the contents beside it link to each section on GitHub.
BugEzy 平台開發規範
平台概覽
亞洲專屬平價 MCP 語音除錯工具。開發者用中文語音錄 Bug → 自動產出 DOM 軌跡 + Console + Network + 中文字幕 → 一鍵分享 → AI 透過 MCP Pull 模式直接讀取修復。 FOX 是唯一創辦人(同時也是 LottoShare 創辦人),外送員 + 雙專案創業者。
角色分工
- Claude Chat(Opus) = PM/策略,出任務、驗收、維護文件
- Claude Code(Sonnet,在 Zed 編輯器) = 工程師,讀 job 檔執行
- FOX = 最高管理員 + 手動驗收
七代迭代計畫
| 代 | 名稱 | 狀態 |
|---|---|---|
| ① 能錄能存 | Chrome 擴充骨架(rrweb + Console + Network) | ✅ 完成 |
| ② 能聽能說 | Web Speech API 中文語音辨識 | ✅ 完成 |
| ③ 能存能看 | Cloudflare Workers + Supabase + R2 + React 報告頁 | ✅ 完成 |
| ④ AI 能讀 | MCP Server(Pull 模式 12 Tool,回應附 token 省錢對比)= MVP 封測 | ✅ 完成 |
| ⑤ 能收錢 | Google OAuth 登入 + 產品首頁(含聯絡資訊)+ 隱私政策頁 + 用量限制 + 兩層定價 + 使用指南/FAQ 頁 + Web Store 文案/zip + 綠界 ECPay 金流(測試環境跑通:單次 + 定期定額月訂閱 + 取消訂閱,CheckMacValue 對官方測試向量驗證)(PM-61~75)→ Chrome Web Store 已送審 + 綠界補件已重送(2026-06-29);換正式 key + 等審核通過 | 🔨 進行中(PM-75,待審) |
| ⑥ 更好用 | 六模式(錄製/⏪回溯/截圖/🔍即時監控/🔇鍵盤/🖥終端機CLI)+ 跨頁錄製 + 時間軸標記 + AI精簡/校正 + 乾淨toggle;上架前打磨(跨頁游標/CSP 相容/語音穩定,PM-68~70) | ✅ 完成(PM-27 |
| ⑦ 規模化 | 日韓越語音、跨境除錯鏈、企業自託管 | 待做 |
技術架構(第 2 代完成後)
Chrome 擴充(Manifest V3 + TypeScript + esbuild)
├── inject.ts(MAIN world)
│ ├── rrweb DOM 側錄
│ ├── Console 攔截(warn/error)
│ ├── Network 攔截(fetch/XHR 4xx+5xx)
│ └── SpeechRecognition 語音辨識(zh-TW)
├── content.ts(ISOLATED world)— postMessage 橋接 + storage
├── background.ts(Service Worker)— 狀態管理 + Badge
└── popup.ts + popup.html — 三態 UI + 複製/匯出 JSON
專案路徑
- 主專案:
C:\dev\bugezy - 接手指南:
docs/BugEzy_project_status.md(270 行,§1~§11 完整全貌)
技術棧
- Chrome 擴充:Manifest V3 + TypeScript + esbuild
- DOM 側錄:rrweb 2.0-alpha(JSON,非影片)
- 語音辨識:Web Speech API(inject.ts MAIN world + 頁面授權橫幅)
- 語音降級:Groq Whisper API(待第 6 代)
- 後端 API:Cloudflare Workers(✅ 已部署 bugezy-api.bugezy-api.workers.dev:reports + /mcp + /api/summarize)
- 資料庫:Supabase PostgreSQL + Auth(✅ reports 表上線;Auth 待第 5 代)
- 檔案儲存:Cloudflare R2(✅ rrweb + screenshots 大檔)
- Web 報告頁:React + Vite(✅
@rrweb/replay回放 + 四面板 + 深色主題) - MCP Server:stdio(
mcp-server/)+ Workers/mcp(12 Tool,加get_live_errors/get_terminal_logs/get_screenshots/get_usage_stats;每次回應附 token 省錢 footer)(✅ 已驗證) - 截圖標注:annotate.html canvas(畫筆/箭頭/框/文字)+ 三擷取模式 + 即時字幕(✅ 第 6 代)
- AI 精簡:Cloudflare Workers AI
@cf/meta/llama-3.3-70b-instruct-fp8-fast(✅ 已部署) - 跨頁錄製:inject 即時 flush → background
chrome.storage.localbuffer → STOP 合併去重(PM-34~37,頁面跳轉不丟資料) - 編輯頁:停止後 edit-report(
@rrweb/replaymini player + 時間軸標記 📌 + 乾淨/原始 toggle + AI 精簡),markers 經 MCPget_report_overview可讀 - ⏪ 30 秒回溯:inject 背景循環緩存(不必先按錄製),按⏪打包最近 30s(PM-50)
- 🔍 即時監控:背景每 10s 推 live errors → R2 暫存 → AI 經 MCP
get_live_errors查當前頁,不產報告、token 極低(PM-51/52) - 🖥 終端機 CLI:
cli/npx bugezy-watch -- <command>攔截 stderr/throw/crash → R2 → MCPget_terminal_logs(PM-53) - 測試專頁:server
GET /test+ page2/3 +/test/api/:status(可預測 Bug 場景,PM-48) - 即時監控/終端機暫存:Cloudflare R2 單一物件(非全域 Map — 跨 Worker isolate 才一致)
- Token 省錢透明度:MCP 每次回應附 token 估算 + 對比 Claude in Chrome 省錢 %(PM-54);edit-report 上傳前各區塊估算(PM-55);Supabase
mcp_usage月度統計GET /api/usage/monthly+ MCPget_usage_stats(PM-56,記錄要 await 否則 Workers 提前終止丟寫入) - 報告頁:Server 直接 serve
GET /report/:idHTML(深色 DevTools Tab 分頁 + Token,vanilla JS 讀/api/reports/:id,PM-59);web React 版ReportPage同款 Tab(PM-58,Worker 未服務 SPA,靠 Vite 預覽) - AI 語音:精簡
/api/summarize(Llama 3.3,轉重點)+ 校正/api/correct(Llama 3.3,修錯字/去贅字/還原術語、保留原意,PM-60c)。踩雷:測中文 API 用 Python 送 UTF-8,別用 Windows Git-Bash curl(會把 body 編碼壞 + 終端機顯示花掉) - 登入:Google OAuth(
chrome.identity.getAuthToken+ manifestoauth2/identity)→POST /api/auth/google(Google userinfo 驗 → Supabaseusers)→ session 存 storage;報告帶user_id(PM-61,第 5 代前置;擴充 ID 須對上 Google Console 註冊值) - 對外頁面(上架/審核用):產品首頁
GET /(HOMEPAGE_HTML,Hero + 六模式 + 兩層定價 + 聯絡 email,PM-62/65)+ 隱私政策GET /privacy(PRIVACY_PAGE_HTML,中英雙語,PM-64)+ 使用指南GET /guide(四步驟 + 六模式 + MCP 設定)+ FAQGET /faq(手風琴 14 題,PM-66)。皆深色主題、無 JS、RWD;首頁 footer 串四頁。Chrome Web Store 上架文案存docs/chrome-web-store.md(PM-67,權限說明依實際 manifest 校正)。綠界 ECPay 已送審 - 上架前打磨(PM-68~70,皆待瀏覽器實機驗收):跨頁回放游標(edit-report
injectCrossPageCursor每段新 FullSnapshot 後補合成 MouseMove);CSP 相容(注入早是宣告式world:MAIN免改,inject.ts 把頁面 MAIN world 的innerHTML改 DOM+textContent以過 Trusted Types 如 GitHub);語音穩定(inject.ts 補onstart重置計數、onerror 分類 no-speech/audio-capture/權限/其他、統一狀態指示 🟢🟡🔴) - popup 版本更新通知(PM-71):
checkVersionNotice()用chrome.storage.local的bugezy:lastVersion比對 manifest 版本,有舊記錄且不同才跳.update-notice卡片(首裝不跳)。Web Store 打包:PowerShell + .NETZipFile.CreateFromDirectory從dist/取執行期 11 檔(排.map)→bugezy-v0.1.0.zip(manifest 在根);offscreen/icons 實際不存在故不含 - 綠界 ECPay 金流(PM-72/72b/73,測試環境跑通):
GET /checkout?user_id=回自動提交綠界表單(含定期定額 Period 參數,月扣 NT$80)→ 付款 →POST /api/ecpay/callback(首期)/POST /api/ecpay/period-callback(第 2 期起)驗 CheckMacValue → 更新users.plan(成功 paid/失敗 free,並記ecpay_trade_no+ 展延plan_expires_at)→POST /checkout/result結果頁。取消訂閱(PM-73):POST /api/user/cancel呼叫綠界/Cashier/CreditCardPeriodAction(Action=Cancel + TimeStamp,非/CreditDetail/DoAction) → 標plan='cancelled'(用量檢查視同 paid,到期由/api/user/plan自動降 free);popup 顯「取消訂閱」/「已取消,可用到 YYYY/MM/DD」。plan三態:free/paid/cancelled。popup 升級鈕開/checkout。CheckMacValue 踩雷:依綠界官方 AI SkillECPay/ECPay-API-Skill(guides/13+test-vectors;端點/欄位看guides/01)——TS 的encodeURIComponent不編碼~/',需額外~→%7e、'→%27、%20→+再小寫 + 還原 .NET 7 字元;Workers 無同步 crypto 改用crypto.subtle(async SHA256)。務必對官方 8 個測試向量驗證。測試帳號 MerchantID 3002607(公開),正式上線換正式 key(HASH_KEY/IV 建議wrangler secret)。schema:users加ecpay_trade_no/plan_expires_at(FOX 跑 ALTER) - 用量限制(PM-63):
FREE_LIMITS(錄製 10/回溯 5/MCP 20 月)+getUserIdFromHeader(解Bearer base64(user_id:ts));GET /api/user/plan(查方案 + 剩餘 + 跨月自動重置)、POST /api/user/usage(遞增,免費版達上限回 403limit_reached、付費版 unlimited)。popup 顯示「剩 N 次/已用完」+ 升級提示;background 錄製前checkRecordingUsage()(未登入/API 不通不擋)。schema 加users.recording_count/rewind_count/mcp_count/usage_reset_at。目前僅 recording 串前端,rewind/mcp 後端就緒待接
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 146 lines · 101 tokens per session scan A 2f3de5dd070d
BugEzy MCP Voice Debug Tool Dev Guide is a skill published in the GitHub repository fox100039-design/bugezy (0 stars, last pushed 2d ago), licensed MIT. It adds 101 tokens to every session and 4,296 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
debug-optimize-lcp
Guides debugging and optimizing Largest Contentful Paint (LCP) using Chrome DevTools MCP tools. Use this skill whenever the user asks about LCP performance, slow page loads, Core Web Vitals optimization, or wants to understand why their page's main content takes too long to appear. Also use when the user mentions…
systematic-debugging
Use when debugging a failing test, build error, or runtime issue that isn't immediately obvious. Guides a 4-phase root cause analysis instead of random fix attempts.
diagnose
Trace from a reproduced symptom to the source code that causes it. Pin the specific file and approximate line, rate confidence in the cause and clarity of the fix independently, and always propose a concrete fix.
repro-admin
Reproduce an EmDash admin UI bug. Attach a container, start the demo dev server, drive the admin with agent-browser using the dev-bypass session, and capture the reproduction as screenshots plus a replayable transcript.
log-error-digest
Analyze log files to troubleshoot errors, identify peak error periods, and produce error clustering, frequency statistics, and time distribution reports. Supports JSON, syslog, and Nginx formats with automatic detection. Use when a user uploads a .log file and asks to analyze errors, find patterns, debug issues, or…
byted-util-volcengine-detect-retry
An orchestration workflow for Volcengine Cloud Detect, a service that checks websites or network endpoints from test locations.