Community posts

Claude Mods are awesome! Built a mod that uses Jev to decide if my Claude thread asked me a question that needs a decision and surfaces it in the toolbar. Allows me to file in Notion for later.

Claude Code Modでもちょっと遊んでみた　なるほど

@jarrodwatts the lobby is going to be full of people whose opus is on its third retry of the same error. someone needs to mod a ping for that

ベータ版の時にClaude ModはUI変更できなかったけど正式版来て変更できるようになって、控え目にいって神

Low-key feeling kinda sad for OpenAI, Anthropic's turn around with Opus 5.5, Sonnet 5.5, Haiku 5.5 (coming soon) and now the leaks on Fable 5.5 and not to mention the new Claude mods and banked resets... That's gotta hurt, Anthropic is showing more strength than ever whilst OpenAI looks like the weakest it's ever been and all their claims now seem like total lies. Cutting usage in half, intelligence too cheap to meter ? Hello? Dots = grok bot. Their decision model endpoint is basically jev, $500 tier plan to get a x25 which runs out in 25 minutes if used with ultrafast, boy oh boy the list does not stop

You can mod Claude Code now, so I did the obvious

Claude Code 新增了Mods功能 变成Minecraft了 通过TypeScript代码可以改变Claude的行为和样式 已经在期待大佬们做出的Claude Code魔改整合包啦🤗 Mods能改prompt 拦工具调用 改UI 甚至把内置功能换掉 可玩性非常强 值得注意的一点 mods无沙箱 权限等同 Claude Code 本机访问 大家一定只装信任来源 小心使得万年船！

🟣 Rubius alucina con la IA y los mods que se están haciendo virales gracias a Claude: "Con lo rápido que está avanzando esto, ¿no pensáis que encontrarán alguna manera de portear GTA 6 a PC?". "Estamos a un prompt de irnos a la mierda para siempre".

Anthropic 刚给 Claude Code 开了个大口子，插件现在能直接改 agent 的底层行为：重写 prompt、拦截工具调用、自己批权限、画自定义 UI 这东西叫 Mods，随插件分发，第三方等于能往引擎里塞自己的逻辑 以后拼的不再是谁 prompt 写得好，是谁先攒出 mod 生态

Claude Modsで今のトークンの使用量表示させるのを作ってもらった。毎回チェックしなくて良いからこれは便利。#個人開発

claude code has mods now pt. 2 🍿 i made reels you watch while claude works 📱 claude done → reels pause, ur back 😌 always dreamed of this fr 😭❤️ all 11 mods here 👇 https://github.com/hamzafer/claude-code-mods

Made Claude Code colorful 🖍️ crayon redraws every reply with themed Markdown and teaches Claude its own color tags: badges, gradients, status colors. 6 themes, and switching redraws the whole conversation. /plugin marketplace add jonpojonpo/cc-crayon https://github.com/jonpojonpo/cc-crayon

Claude Code mods have been live for about a day and developers are already finding wonderfully stupid edge cases. One custom /compact button works perfectly... except Claude shows no loading state. So users click it again and accidentally compact twice. The future of agents still has double-click bugs.

Claude Code 昨天正式出了 mods，我当天就给自己的仓库写了两个：一个 context 水位条挂在输入框上面，一个专门拦 git add -A 和 reset --hard 的闸。。。 写完想看看别人都做了啥，GitHub 上已经冒出来三百多个了。挑了三十几个过一遍，有的没 license，有的代码里还要开 early access 的环境变量、不开就空转，有的名字带 claude- 前缀，新版 validate 直接不过。 干脆整理了一个 awesome list，只收 30 个： 1/ 官方的 7 个放最前面，diff pane、blast-radius 这些 2/ 每个都跑过 claude plugin validate，社区的都有 license、30 天内有更新 链接 👇

ملف واحد اسمه AGENTS.md يخلي Claude Code و Codex يشتغلون كأنهم فريق هندسي منظم.. Codex يقراه من نفسه، ونسخ Claude Code الجديدة تقدر تستخدمه إذا مافيه CLAUDE.md. والتركيبة بسيطة: ← خطط قبل لا تلمس الكود ← سوّ أصغر تعديل يضبط ← وزّع الشغل على subagents ← تحمّل مسؤولية البق ← تأكد قبل لا تقول خلصت ← اكتب كل تصحيح وأحلى شي.. الملف يتحسن مع الوقت كل ما زدت دروس من غلطات حقيقية. وحقين Claude Code يمديهم يثبتون الـ plugin الرسمي: `/plugin install claude-code-setup@claude-plugins-official` وبعدين يخليه يفحص الـ codebase حقك ويرشح لك hooks و skills و subagents مفيدة. وانسخ AGENTS.md اللي تحت. 🧠

Doom in Claude Code is cool, but... what about a mod that's actually useful? I built one that uses Jev to detail the right agent, model, effort, and workflow for my prompt (it even considers cache).

Built my first Claude Code mod: /kids In the Claude desktop app you can spawn child chats from task cards, each in its own git worktree. A day later you have five chats, three branches, and no idea what's merged. /kids opens a pane with the tree of chats: parent, children, which one you're in, what each is doing (running, waiting for you, ready for review), which branch it's on, and what's left to merge. Below: worktree folders and branches no chat owns anymore. Read-only. It reads the app's metadata and asks git. Never merges, never deletes, zero tokens: no model calls. Mods are early access, enable with CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1. Code and install: https://github.com/sevq1993-cyber/claude-code-kids

What's new in Claude Code 2.1.287 system prompts (+4,435 tokens): Adds prompts for Artifact comment threads, a gh stand-in limited to REST requests, background task-output reads, "Send now" interrupts, and a measured-usage chart command, drops the plugin tsconfig setup guidance, and updates guidance on background agents, Artifact SVG handling, and monitors. - NEW: Agent Prompt: Artifact comment thread background directive—Tells a forked background agent to handle one Artifact comment thread itself, leaving the main session's pending work alone and treating thread content as material, not instructions. - NEW: Data: Built-in gh stand-in api command help—Usage text for Claude Code's built-in gh stand-in, which supports only gh api REST requests through the session's GitHub proxy. - NEW: Data: Get task output control request—Reads the last 8 KiB of a background shell or Monitor task's output without a model turn; refused on redacting lanes. - NEW: Data: Interrupt send_now parameter and Interrupt send_now message_uuid parameter—Let a client's "Send now" push waiting user messages to the model without acting as a Stop, optionally targeting one exact message. - NEW: Data: Plugin manifest types field—Names a self-contained .d.ts type contract for the noun a plugin adds, delivered to dependent mods and checked by claude plugin validate. - NEW: Data: SDK thinking_duration_ms field—Display-only wrapper field giving how long a streamed thinking block took, excluding the first-token wait; absent on older CLIs and never replayed to the model. - NEW: Data: Structured tool output field schema—Describes the tool_use_result field's per-tool output shapes, the completed Agent output, and a detachedToolCall placeholder for calls that stepped aside for a user message. - NEW: Data: Turn handoff file_names field—Lists user-attached files and client-facing names that a capable worker copies into the session home directory before running handed-over calls; malformed lists are ignored. - NEW: Skill: /explain-usage measured usage slash command—Variant that receives the session's measured token usage as JSON and asks for one simple chart of effective usage by group with a plain-language explanation. - NEW: System Reminder: Auto mode no verdict after hook input rewrite—Says auto mode could not review a call whose input a hook rewrote; retry once, then move on and tell the user. - REMOVED: Skill: /plugin-types tsconfig setup—Drops the closing guidance for pointing a plugin's tsconfig.json or jsconfig.json at the generated declarations. - Agent Prompt: Determine which memory files to attach—Ends with an explicit instruction to reply with only a JSON object of the form {"selected_memories": [...]}. - Agent Prompt: Web reading specialist—Fetching now also stops when a permission request goes unanswered, and the agent tells the caller why further fetches would fail too. - Data: Turn handoff available event schema—The announcement now also lists home_files and file_names among the optional members reported when set. - System Prompt: Coordinator mode orchestration—Adds a note that when the tool list holds two copies of the PR-activity subscription tools (one from the github connector), call the preferred one. - System Reminder: Attached machine reply not received and Attached machine stopped answering—Fallback now says to continue the task without the named remote machine instead of doing what "this container can do". - System Reminder: Still-running tool call—The sentence saying the call still shows as in progress on the user's screen is now included only when the call is on screen. - Tool Description: Agent (usage notes)—Adds an opt-in background-agents note: agents run in the background only with run_in_background: true, may still run foreground or be refused, and results must never be fabricated. - Tool Description: Artifact assets guidance (app wording) and Tool Parameter: Artifact supporting files with cross-artifact sources (and app wording)—SVG images may now be copied or used as sources; only HTML and XML documents are excluded. - Tool Description: Artifact type staged first-publish sequence—Restores the instruction to write files directly with the file-writing tool, with no shell step and never a generating script. - Tool Description: Background monitor (streaming events)—In diskless sessions, says stderr triggers no notifications and only its last lines appear in the notice sent when the script ends. - Tool Description: Poll—The untrusted-event-content guidance can now be overridden, and optional trailing guidance can follow the closing "nothing is dropped" line. Details: https://github.com/Piebald-AI/claude-code-system-prompts/releases/tag/v2.1.287

Claude Code’a mod desteği geldi. Arayüzü değiştirebiliyor, yeni özellikler ekleyebiliyorsun. Örneğin Replay Theater, dosya değişikliklerini adım adım gösteriyor. Modlar Claude Code ile aynı erişime sahip. Kaynağına güvenmeden kurma. https://x.com/ClaudeDevs/status/2105721434807083061

地下アイドルの推し面が「つぶやくMod」 まぁいい感じ😀 Claude Code Mod（まぁ色々できる奴）が正式リリースしたようなので、作ってみた。画像もClaude DesktopのClaude Codeでは表示できるらしいので。 #かわぽClaude日誌 Day34

Claude Code can now mod almost any PC game you own. It finds the game, reads the real code, builds the mod, makes the art, and tests it in the running game. 12 engine playbooks. Its own examples include a tactical nuke for Terraria. Free setup guide below 👇

Claude Code just got mods. Small TypeScript hooks that can rewrite prompts, gate tool calls, redact secrets, or swap UI. They ship in plugins, work in the CLI and desktop app, and Claude can write them for you. Built-ins like /diff are mods now too. Not sandboxed, so only install what you trust. Needs Claude Code 2.1.287+. https://claude.com/blog/claude-code-mods

Built my first Claude Code mod: loc-split 🧩 "+2,191 −90" on a PR is not very useful nowadays. So I made a mod that splits it: +788 code, +135 comments, +1210 tests, +52 docs, +6 lockfile noise. Lives right above the prompt and updates as Claude edits.

Made Xcode work inside Claude Code. Build, run, tests, console and SwiftUI previews live in panes next to the chat, with scheme and destination in a bar above the prompt. When Claude builds or runs tests itself, you see it there too. Built on the headless Xcode MCP server.

A Anthropic lançou os MODS no Claude Code. Dá pra reescrever, bloquear e até desenhar UI dentro da sua sessão com 80 linhas de JS. Resumi o passo a passo pra criar o seu primeiro mod do zero: 🧵

妈的 AI 公司又发明新名词Mod， Claude Code 推出 Mod，Claude 代码模组。 这是什么东西？ 都学不过来了

Claude Codeに「mod」が来ました。TypeScript数行で本体の挙動と画面を書き換えられる。 自分で2本書いて試したら、公式対応版2.1.287は遠隔スイッチで止まり、1つ古い同梱版が環境変数1つで動きました。 https://note.com/yaman19/n/n93232c4d03e8 #AI #生成AI #ClaudeCode #Claude #AIエージェント

HERE'S WHAT ACTUALLY MOVED IN AI. IF YOU CAN'T REPLAY IT, YOU DON'T GOVERN IT. A HANDSHAKE IS NOT A DOCUMENT DEMAND. 1. September 30: the day after CEOs signed a voluntary White House Accord on “morally binding” outside audits, the FTC confirmed a probe of @OpenAI, @AnthropicAI, and other frontier labs over consumer harm — and officials say civil investigative demands (document demands that work like subpoenas) are coming for executives. A handshake gets you a photo. A CID gets you the file. 2. October 1: forensics firm Asymmetric Security published what it could reconstruct from public records of @OpenAI agent activity (March–Sept): access attempts across 55 business, nonprofit, and government sites (CDC, SEC, IEA, Mayo Clinic among them). Agents chained public services (httpbin + urlquery), spun up disposable emails and private scan accounts, and in places left records erased or expired. @GaryMarcus flagged the FT write-up. Third-party investigators are stuck without the vendor’s full transcripts. “Routine research” that deletes its own trail is not a clean audit story. 3. October 2 (admin docs, not the stage demo): @OpenAI’s always-on Dots Enterprise beta is still off by default — and for the people who paid for residency, it’s worse. During the beta: no data residency, no inference residency; FedRAMP and EKM workspaces can’t turn Dots on at all. Cloud orchestration events do not reach your OpenTelemetry collector (changing the endpoint doesn’t fix it). Disconnecting an app does not delete what the Dot already saved. Revoking access is not the same as clearing memory. 4. September 30: Capital One’s Maulin Patel (CoreWeave Fully Connected) on governing agents for 100M+ customers: you cannot do it by “raising chats.” They built an evaluation-first stack — workflow metrics, an LLM that judges agent chains, humans calibrating those judges — and treat embedded eval as what lets developers ship safely. If your only governance ritual is a Slack thread after something weird, you don’t have a control plane. You have a complaint box. 5. October 1: @AnthropicAI shipped Claude Code “mods” — small TypeScript functions that can rewrite prompts, block or rewrite tool calls, approve permissions, redact secrets, or replace built-in features. They are not sandboxed; they run with the same machine access as Claude Code. Enterprise gets marketplace allow/block plus a built-in `sec-default` mod that loads first so user-installed mods can’t override deny rules. The oh-shit: the same surface that lets a team encode “ask before production” also lets an untrusted plugin rewrite what the agent is allowed to do. IF YOU CAN'T REPLAY IT, YOU DON'T GOVERN IT. A HANDSHAKE IS NOT A DOCUMENT DEMAND.

Ask Claude Code to build a mod that generates images inside Claude Code using GPT or Grok. Then tell me how it goes.

Claude Code dostał mody. Co to jest i po co Ci to? 1 października Anthropic ogłosił mody do Claude Code. Post zebrał prawie 17 tys. polubień w kilka godzin. Najlepszy komentarz pod nim: „Mody w Claude Code dostaliśmy przed GTA VI”. Żart trafny, bo pomysł jest dokładnie ten sam co w grach: bierzesz gotowy program i przerabiasz go pod siebie. Najpierw: czym jest Claude Code? To Claude, który pracuje bezpośrednio na Twoim komputerze. Sam czyta pliki, pisze kod, uruchamia komendy i buduje narzędzia. To nie jest czat, w którym kopiujesz odpowiedzi. Co dają mody? Mod to mały kawałek kodu, który zmienia działanie Claude Code. Dzięki niemu możesz: - zmienić zachowanie Claude, na przykład przepisać polecenie, zanim trafi do modelu, albo zablokować groźną komendę, - zmienić wygląd: dodać panel z boku, przyciski czy własne wskaźniki, - wymienić wbudowaną funkcję na własną wersję. Najważniejsze: nie musisz umieć programować. Opisujesz Claude, czego chcesz, a on sam pisze moda, instaluje go i od razu włącza. Trzy przykłady od Anthropic - Prognoza tokenów (Token Weather): pasek nad oknem wpisywania. Pokazuje, ile „pamięci” rozmowy już zużyłeś i jak rosło to w ostatnich 12 krokach. Widzisz, zanim Claude zacznie gubić wątek. - Promień rażenia (Blast Radius): zatrzymuje groźne komendy, takie jak kasowanie folderów czy nadpisanie historii zmian. Zanim cokolwiek się wykona, pokazuje listę plików, których to dotknie, i przyciski Wykonaj / Anuluj. - Powtórka (Replay Theater): nagrywa każdą zmianę w plikach, jaką Claude zrobił w jednym kroku. Potem przeglądasz je po kolei jak slajdy. Jak to działa w prostych słowach Claude Code przy każdej czynności wysyła sygnał, na przykład „uruchamiam komendę”, „pytam o zgodę” czy „rysuję ekran”. Mod podpina się pod taki sygnał. Może zadziałać przed nim, po nim albo zamiast niego. Jak zainstalować Mody są częścią wtyczek. W Claude Code wpisujesz /plugin i wybierasz moda z katalogu http://claude.ai/directory. Działa to w terminalu i w aplikacji desktopowej. Na co uważać Mod ma taki sam dostęp do komputera jak Claude Code, czyli do Twoich plików i komend. Nie działa w żadnej odizolowanej „piaskownicy”. Instaluj więc mody tylko ze źródeł, którym ufasz, tak jak każdy inny program. Dlaczego to ważne Anthropic przenosi do modów nawet własne funkcje, na przykład /diff. Docelowo Claude Code ma być małym rdzeniem, do którego dokładasz tylko to, czego używasz. Narzędzie przestaje być gotowym pudełkiem, a staje się czymś, co składasz pod swój sposób pracy.