Community posts

Claude Code 刚开的 Mods， 已经被玩疯了。 10 月 1 日 Anthropic 上线的功能： 几行 TypeScript， 就能改写 Claude Code 的内部行为—— 拦截危险命令、给输出脱敏、重写提示词。 上线这才几天， 社区已经盘出了"10 个最值得装的 mod"： 看图、监控子 agent、省缓存钱， 甚至等 Claude 干活的时候打 Doom。 连 DeepSeek 都坐不住了， Harness 昨天更新， 直接加了 Claude Code Mods 兼容层。 看懂这个信号： AI 编程工具正在从"工具"变成"平台"。 当年 VS Code 靠插件生态干掉一票编辑器， 现在剧本重演—— 谁拥有扩展 API，谁就拥有开发者。 你在用 Claude Code 吗？ 装了几个 mod？

Loving the new mods in Claude Code. My first: a Handoff button. One press saves where you are, clears the chat and continues with a fresh context window. Super quick, super easy, nothing lost. Built on @mattpocockuk's skills

Anthropic 发布的 Claude Mods，通过 JS / TS Hooks 可以直接介入 Claude Code 的运行流程： → 重写 Prompt → 拦截/修改工具调用 → 添加自定义 UI → 替换内置功能 → 给 Agent 增加新的工作流 Mods 会打包在 Plugin 里，可通过 /plugin 安装，并支持 Claude Code CLI 和桌面端。 但有个风险需要注意： Mods 不在沙箱里运行，权限基本和 Claude Code 本身一致，只装可信来源 以前是在配置 Claude Code；现在已经可以开始“改造 Claude Code”了

@ClaudeDevs Plugin que filtra o output do Claude para 'info importante que podes perder' é só uma attention layer em cima de um modelo que já falhou em sinalizar. Curioso com a precisão: quantos flags são load-bearing vs ruído? Alguém tem false-positive rate? 🔗

こ、これが Claude Code mods... ｺﾞｸﾘ....

@dani_avila7 Hola, @dani_avila7! me encanta esto. En algún momento intenté ponerlo dentro del script de statusline pero quedaba estático siempre... Así que esto resuelve totalmente mi idea. Me gustaría saber si podemos tener un setting para que quepa en una sola línea, en especial en el CLI

I built 12 Claude Code mods (all free). So you never ask Claude the same thing twice: Access them all here → https://charliehills.substack.com/p/claude-code-mods Anthropic literally just dropped mods for Claude Code. A mod changes how it looks and behaves. Here are my favourites (so far): Watch the work ☑︎ Mission Control → the plan and progress, live ☑︎ Open Loops → every ask, ticked only with proof ☑︎ Sessions Band → all your open sessions, one band ☑︎ Office → a pixel-art desk for every session ☑︎ Done Ping → a sound when Claude finishes Keep it safe ☑︎ Safe Delete → deletes go to the Bin, with undo ☑︎ Secrets Guard → Claude never sees your API keys ☑︎ Outbox → every message waits for your Send ☑︎ Pre-build Check → re-reads your rules first Better answers ☑︎ Show It → opens what Claude just made ☑︎ Plain Reply → flags long, jargon-heavy replies ☑︎ Mod Maker → spots your repeat asks, builds the mod Repost ♻️ to help someone in your network.

@Joshua_WD 中文版来了 🦀 给 Claude 桌面 App 养了只像素桌宠 Clawd：用量一眼看完，Claude 干活它也干活，额度告急先替你流汗，缓存命中一高就戴墨镜摸鱼。开源免费，两行命令领养 👇 https://github.com/manson341349-beep/claude-desktop-mods

Update: Clawd 1.2.1 🦀😎 · English UI, follows your system language · Chill mode: cache hit ≥ 90% → sunglasses on (idea by @Joshua_WD) · …and he takes them off now and then Update with: claude plugin update usage-pet@claude-desktop-mods

Claude Code Mods are pretty cool

IF YOU RUN CLAUDE CODE IN BYPASS MODE, READ THIS hamzafer made a pack of free claude code mods. one of them, blast-radius, holds rm -r, force pushes and migrations until you say ok i ran it in a sandbox with bypass permissions on and told claude to rm -rf a folder. it still stopped and showed what would get deleted. nobody approved, so it cancelled. folder survived the pack also has mission-control, a live map of every subagent you have running bypass mode without something like this is how you lose a weekend

Claude Code 上线 Mods：几行 TypeScript 就能改写它的内部行为——拦截危险命令、给输出脱敏、重写提示词。 一人公司用法：让 Claude Code 自己写一个 Mod，部署前强制二次确认。 它没沙箱，权限和本体一样：只装信任来源的 Mod。 #ClaudeCode #AI编程 #独立开发

Claude Code can now be modified from the inside. Mods can react to prompts, intercept tool calls, change behavior and even alter parts of the interface. We went from: AI coding tool → programmable AI coding environment. But Mods aren't sandboxed. Would you trade safety for that level of extensibility?

Analyze the way I currently use Claude Code. My typical workflow is: [Describe your workflow] Find the parts that are: - Repetitive - Time-consuming - Difficult to manage - Easy to automate - Frequently repeated - Limited by the default experience Give me 10 potential ideas for Claude Code Mods. For each idea, explain: 1. The problem 2. Why it matters 3. What the Mod would change 4. How it would work 5. How often I might use it 6. How difficult it would be to build Finally, rank the ideas from highest value to lowest value.

Imagine telling Claude: “Build a feature for yourself.” Now you can. Claude Code Mods let you customize the experience with your own features, behavior, UI, and workflows. Instead of only using Claude Code… You can start building the version that fits your workflow. Here’s how:🧵

做了一個在桌面版工作列旁顯示Clawd動態的mod。 顯示在最上層，就算跳去其他視窗也能看見 這下就算不用停留在Claude Cose內也能知道目前的工作狀態了~☺️

10 个好用的 Claude Code Mods ，你知道几个？ 很多人不知道 Mod 是什么？ 它和我们之前 Plugins 有什么本质区别？ 你可以将 Plugins 理解成接入外部的能力，而 Mods 是 Agent 的「自我改造」， 它可以修改 Claude Code 的原来 UI 和原本行为。 用户的操控自由度变得越来越随心所欲啦。

AIニュース動画の改善版です！ ElevenLabsの音声＋ローカルのLTX-2.3で映像を再作成。 Claude Codeの「Mods」で、作業状況の表示や操作を拡張。使い方と注意点を男女の掛け合いで約85秒にまとめました。 ※AI生成映像・音声。出典は返信へ

Claude Code mods let developers add live context readouts and custom review panes inside a coding session. The October 1 tutorial builds Token Weather, a context-window monitor, and shows Replay Theater for stepping through edits. It requires Claude Code 2.1.287+. Mods run with Claude Code’s access, and the API can change between releases; inspect the publisher’s code before installing. https://claude.dev/blog/getting-started-with-claude-code-mods/ #ClaudeCode #TokenWeather #ReplayTheater

Claude Code Mods finally clicked for me hqmank made an animated diagram that follows your data as a mod steps in. docs could never (via @hqmank)

我重新做了我的Github Readme 并且提前把lemo-wake开源出来了。一周开源了三个项目，也是得益于claude 这波reset，开了三个号三个reset 一共6个周限。 Lemo-opuscar 一个claude视频制作风格库，现有43种风格，提供风格导演剧本指南等提示词并提前搭建工具环境。 Lemo-mod 一个给claude code 的换装mod，20个风格，cli和app兼容。同时提供一系列安全的mod功能模块，按需开启。 Lemo-wake 一张图片，照片，聊天截图等，变成5秒GIF，支持输出mp4，gif，webp，livephoto等，花样百出。 欢迎大家到我的主页玩。

本周收藏这期质量可以，AIHOT 自己抓热点写日报、universal-modder 让 Claude 给单机游戏打 mod，三个都值得 mark

This animated diagram shows how Claude Code Mods work. Follow the data and see what a mod does inside a Claude Code session.

I highly recommend using Cache Control in Claude Code 🦀 This is probably one of those mods that helps you save tokens the most and keep sessions alive much longer. He adds a separate strip for Claude's 5-minute cache and sends notifications to help keep the cache warm. The command to install: npx claude-code-templates@latest --mod observability/prompt-cache-control

Claude Codeのmod機能でCodexのペットを輸入！ついでに使用量も体力ゲージ風に。捗る～(*´ω`*) #ClaudeCode

Whoever owns the extension api wins the harness fight. DeepSeek Harness clearly thinks so: it shipped a compat layer for Claude Code mods less than two days after Anthropic launched them on Oct 1 It's experimental, and the release was signed off by DSH's lead, Tianyi Cui. His take: in Claude Code, mods are just plugins. DSH was built as plugins all the way down Smart. That's pretty much how the VS Code extension api became the thing every new editor had to support One catch though. Claude Code's built-in sec-default mod is listed as not runnable in DSH yet. So your mods come along, and the safety default stays behind Anyone actually planning to run their Claude Code mods in another harness?

@ClaudeDevs Control Claude Subs, Local Machine, Virtual Machine, Git Actions and VM on/off. Mods is a small upgrade but adds a lot productivity.

I’ve added 2 Claude Code mods to the collection. The collection now has 28 mods for you to find and use. - claude-clawd: https://kienhoang.me/claude-code-mod/clawd - claude-usage-mod: https://kienhoang.me/claude-code-mod/usage-mod Find the Claude Code mods that suit you personally

外掛具備觀察、改寫、接管能力，能攔截高風險終端指令，並遮蔽敏感金鑰。 剛剛，Claude Code 推出 Mods 功能，一句話魔改自己 https://hao.cnyes.com/post/270664

Today I’m releasing my first mod for Claude Code! .claude/mods/xray xray allows you to see under the hood while Claude is working. Glanceable cards show you how what claude is doing, what is left on its plate, plus a turn-by-turn genome showing the history of the session.