Community posts

@lydiahallie Claude bots group chat mod :)

@Berlinworlds It detects subscription vs API and adjusts the TTL

Claude code mods are pretty cool, you can actually make the terminal look pretty. Images in the terminal 😮

Just for fun, I tested how far Claude Code's /plugin-authoring goes , and it is what I expected. This is a toy. The real takeaway is how easily you can build your own productivity tools right inside Claude Code 👇 Oh,and my context bar is weather now ⛈️ @ClaudeDevs

@lydiahallie, i made a second one! I put live football inside Claude Code ⚽ After Coaster Tycoon, here's my second Claude Code mod: Claude Live Scores. While you work, it keeps the match in the corner of your eye: • A clean, dark scores pane: Premier League, LaLiga, Serie A, Bundesliga and Ligue 1, plus the World Cup, EURO, Nations League and more • A celebration for every goal; red cards flip in and VAR decisions shake • My favourite part: pixel-art Claude mascots in the two teams' kits play the live match right above your prompt. When the real match scores, so do they 🥅 It works in the desktop app and in the terminal, needs no account or API key, and never looks at your code or prompts. The video is rendered from the mod's own code, using the real Spain v Czechia match from launch night. Open source (MIT), and in review for Anthropic's plugin directory: https://github.com/thickiran/claude-live-scores #ClaudeCode #Anthropic #OpenSource #Football #DevTools

/plugin marketplace add netgfx/claude-session-dashboard-mod https://github.com/netgfx/claude-session-dashboard-mod

Claude Codeが大幅拡張、「Mods」登場　内部処理や画面まで開発者が変更可能に https://xenospectrum.com/claude-code-mods-controls

@dani_avila7 Mods are the launch most people skipped this week. Does it show what a keep-warm ping costs vs. letting the cache expire? That's the number I'd want before leaving it on all day.

claude code mods shipped a babysitter for the 5-minute prompt cache. /cache shows 98% hits on 136k tokens and a green bar that yells keep the fire warm

Claude Code mods can approve or deny permission requests. They aren't sandboxed. On Enterprise, a security mod loads first so user mods can't override deny rules. Lesson: guardrails belong in the runtime, not a policy PDF. https://claude.com/blog/claude-code-mods

Claude Code 2.1.287 ships mods: TypeScript hooks that can intercept tool calls and approve permission requests. They run unsandboxed, with the same machine access as Claude Code itself — and they're on by default. So the extension point is also the approval point.

@davekiss @jarrodwatts @ihamzafer Great i have few of my scripts doing similar with twist I will soon be getting a kit for my self as mods from my scripts. Mods has really enabled claude to make more coustom experience with llms after skills. https://github.com/VedantAndhale/claude-pro-kit

@ClaudeDevs Being able to step through file edits should help catch where a run went wrong. I'd want a redaction pass before sharing a replay with someone else. Mods leads today's AI Daily, with the docs linked below it: https://x.com/MisterSnowCrash/status/2106492049587425717

在 Claude Code 里粘贴截图，输入框只显示一个 [Image #1]。 贴了三张图，我经常分不清哪张是哪张，贴错了也看不出来。 这个哥们 Jarrod Watts 写了个小 mod，把粘贴的图直接用缩略图显示在输入框上面。 就这么一个功能，6.9 万浏览，1200 赞，1000 收藏。 评论区一片 FINALLY。 实现上有几个细节挺有意思： 粘贴图片不会触发编辑事件，所以它每 200ms 扫一次输入框，找 [Image #n] 标签。 图片本来就缓存在 Claude Code 的临时目录里，它只读 PNG 文件头拿尺寸。 画图交给终端自己（kitty 图形协议），图片数据压根不经过这个 mod。 不联网，不写文件。 我觉得更值得注意的是 mod 这个词。 Claude Code 从 2.1.287 开始支持 mods，第三方可以直接往它的界面上画东西。 一个命令行工具，开始像游戏一样允许改 UI。 接下来大概就是官方顾不上的小痛点，社区几天就补上，好用的再被官方收进去。 浏览器插件、魔兽的 UI 插件都是这么走过来的。 大家都在卷谁的模型更强， 结果最先火起来的是让你看清自己贴了哪张图。 我已经能预见自己花一下午装 mod，然后一行正事没干。 https://x.com/jarrodwatts/status/2106153410697564235

Anthropic shipped mods for Claude Code — small TypeScript functions that hook the agent loop: rewrite prompts before the model, block/retry tool calls, approve/deny permissions, redact secrets from tool output, swap UI. Built-ins like /diff are now mods you can disable or replace. Team/Enterprise load sec-default first so user mods can't override permission deny rules. Ops read: the agent runtime just became a programmable platform — and an unsandboxed plugin surface. Extensibility without trust boundaries is the new attack surface for agent fleets. https://claude.com/blog/claude-code-mods

AI DAILY 3 October 2026 · Edition closed at 21:02 UTC 01 · Oct 1 · Claude Code Mods let TypeScript plugins change agent behavior, customize the interface and swap features in CLI or desktop. Official demos include context tracking, risky-command previews and session replays. TODAY’S NEWS, REPORTING & RELEASES 02 · David Robinson, who led OpenAI’s safety-report writing, says he quit over the company’s approach to AI risk and calls for aviation-style safeguards. OpenAI says it is strengthening security and outside evaluation. 03 · Sam Altman warned against treating AI as a religious authority or surrendering human judgment to it. 04 · WIRED reports that Meta’s Muse builds profiles of people in users’ lives. Meta says it draws on public information and material users choose to share; users can delete memories or disconnect services. 05 · AI companies need over $4.2T in new revenue within five years to fund the buildout, according to a Bain estimate cited by Reuters. Bain says productivity gains in existing markets alone won’t be enough. 06 · Aleph Alpha released Kolibri-1 with Apache 2.0 weights, German/English reasoning and tool use: 78B parameters, with 3.46B active per token. Native context is 262K, with extrapolation up to 1M. 07 · Higgsfield rebuilt AI Influencer with optional photo-based identity and portrait and full-body views. Characters can move through Genjutsu Motion and be reused in Cinema Studio. 08 · Capcom plans to bring AI into RE Engine development workflows through a gradual roadmap presented October 2. The company has previously ruled out AI-generated assets in finished games. WORTH CATCHING UP ON · OCTOBER 1–2 09 · Oct 2 · GitLab patched a critical AI Gateway flaw that lets authenticated Duo Agent Platform users execute commands. Self-hosted operators should update to 19.2.4, 19.3.2 or 19.4.1; GitLab-hosted gateways are already patched. 10 · Oct 2 · Meta shared six papers developed by mathematicians working with Muse Spark, five reporting answers to open questions. Humans guided and reviewed the work; some problems also have concurrent independent solutions. 11 · Oct 2 · NVIDIA announced a 64GB DGX Spark starting at $4,999, with partner sales scheduled for October 23. 12 · Oct 2 · Cactus released Whistle, a 16.9 MB speech-recognition model that runs offline on CPUs. It supports seven languages and word timestamps, with code and weights available. 13 · Oct 2 · Runway added Seedance 2.5 Draft mode in Agent: generate 480p previews, then enhance them. It is available on all plans. 14 · Oct 1 · Black Forest Labs released FLUX 3 Image with bounding-box composition, targeted edits, up to ten reference images and output up to 4K. Commercial self-hosting requires a weights license. 15 · Oct 1 · Microsoft introduced public-preview streaming transcription in 60 languages with MAI-Transcribe-2-Streaming, alongside MAI-Voice-2.1 and Flash for speech generation in 23 languages. The models are offered through Foundry and MAI Playground. 16 · Oct 1 · Suno’s Speech beta generates spoken narration and its musical backing together. It is available on web and mobile. Follow me for a carefully curated daily AI catch-up. Sources below.

Just got an email from Lydia at Claude Mods are here. Anyone tried them yet?

【TechFeedまとめ】2026-10-04 ■ Zenn 1. Focus Flow AI (v7.0) https://zenn.dev/andy32/articles/7f21569bd64a88 2. LoreKitは本当に「ノーアカウント・ノーネットワーク」か実測してみた https://zenn.dev/mskbhd/articles/lab-304-lorekit 3. MCP・API・Skillsの違いを整理する。AIとの壁打ちで直した3つの勘違い https://zenn.dev/zhaojiayi/articles/d5ad38fc02091a 4. Magnitudeの「llama.cpp比2倍」を整理する — 16GB Macでは逆だった https://zenn.dev/amu_lab/articles/magnitude-vs-llamacpp-16gb-mac-benchmark 5. ローカルLLMにチャートを読ませたら、どの銘柄も同じ答えだった：判断の前に「観察」を聞いて切り分ける https://zenn.dev/charge0315/articles/403143d306c79b 6. 1。「操作者」と「開発者」を分けて罰するAI法案の日、このパイプラインで両方の役割を演じている人数 https://zenn.dev/hideki_tamae/articles/ai-agent-accountability-operator-developer 7. 受注が入ると地球儀にビームが飛ぶ営業ライブボードを、ライブラリなしでClaudeと作った話 https://zenn.dev/arima_dev_lab/articles/35ec34426addb6 8. Claude Codeが、上限に達しても「きりのいいところ」まで続けるようになった https://zenn.dev/shoujiki_panman/articles/claude-code-wrap-up-allowance 9. ChatGPT・Work・Codex・Codex CLIを、作業と確認のしかたで選ぶ https://zenn.dev/sugitakenji/articles/e1819b47908689 10. Codexのモデル／推論レベルの選択、やめました https://zenn.dev/x013318908/articles/c5ac7bf7426b79 11. AIの「全件PASS」を信じない：非LLMの置換スクリプトとHuskyで、AIの書き換え事故を、止まる仕組みで防ぐ https://zenn.dev/souaaiproject/articles/d460978b82ecfe 12. AIへの指示書を別ファイルに追い出したら、最初の入力は8230トークン減った。でも合計は最大1.86倍に増えた https://zenn.dev/numarn/articles/claude-md-pointer-vs-inline-rules-handson ■ Qitta 1. AIを使う前にマクロを使え──先撃ちで表の修正が0.1秒、Geminiと決意した夜の話 https://qiita.com/shu15511551/items/1941cd71b04e8da039d8 2. AIエージェント全史｜スキル標準化からセキュリティ悪夢、PoC地獄を越えて自己改善する時代へ【総集編・2026最新補足つき】 https://qiita.com/kinamocchi_tech/items/754c8dab12da7fa47548 3. AIは資本主義と社会主義の勝敗を決めるのか https://qiita.com/konaito/items/6f7b698ceabf740c850f 4. Claude Code Modsは一番深く拡張できて、一番持ち運べない — 直近のアップデート13個を★で評価（2026/10/3時点） https://qiita.com/faruryo/items/c075c8c3d664c38f2384 5. AIは資本主義と社会主義の勝敗を決めるのか https://qiita.com/konaito/items/6f7b698ceabf740c850f 6. FDEが作る成果物は、ChatAPIであるべきだ https://qiita.com/konaito/items/7c421d21d082f12ec873 7. Claude Code Modsは一番深く拡張できて、一番持ち運べない — 直近のアップデート13個を★で評価（2026/10/3時点） https://qiita.com/faruryo/items/c075c8c3d664c38f2384 8. AIエージェント全史｜スキル標準化からセキュリティ悪夢、PoC地獄を越えて自己改善する時代へ【総集編・2026最新補足つき】 https://qiita.com/kinamocchi_tech/items/754c8dab12da7fa47548 9. AIは資本主義と社会主義の勝敗を決めるのか https://qiita.com/konaito/items/6f7b698ceabf740c850f 10. Claude Code Modsは一番深く拡張できて、一番持ち運べない — 直近のアップデート13個を★で評価（2026/10/3時点） https://qiita.com/faruryo/items/c075c8c3d664c38f2384 11. andraschi/gimp-mcpがClaude Desktopで認識されない → manifest.json書いて手動で拡張機能インストールしたら解決した話 https://qiita.com/beru_ike/items/b8c73be63f5f111008de 12. [Claude Code] ! で実行した後に出てくるメッセージを止める https://qiita.com/murnana/items/072ded7824f825415ec6 13. Windows版ChatGPT / Codexで認証初期化後に消えたProjectsと過去チャットを復旧した記録 https://qiita.com/cal_peace/items/60f3aa23fd27623b2fac 14. V-JEPA2で何ができそうか試してみる written by dots https://qiita.com/daikubo/items/6542ba39a91e0abf7aa6 15. Codexで複数のMarkdown資料から要点・重複・未決事項を整理する https://qiita.com/takuwan_lab/items/beb0a03362a22cbc2924 16. Claude Code Modsは一番深く拡張できて、一番持ち運べない — 直近のアップデート13個を★で評価（2026/10/3時点） https://qiita.com/faruryo/items/c075c8c3d664c38f2384 17. AIは資本主義と社会主義の勝敗を決めるのか https://qiita.com/konaito/items/6f7b698ceabf740c850f 18. FDEが作る成果物は、ChatAPIであるべきだ https://qiita.com/konaito/items/7c421d21d082f12ec873 19. ChatGPTからVS CodeのCodexへコピペせずに指示を渡す――MCPで保存し、人間が確認して送信する構成 https://qiita.com/locolisu/items/81cf8f1140fe58f36e66 20. AIエージェント全史｜スキル標準化からセキュリティ悪夢、PoC地獄を越えて自己改善する時代へ【総集編・2026最新補足つき】 https://qiita.com/kinamocchi_tech/items/754c8dab12da7fa47548 21. andraschi/gimp-mcpがClaude Desktopで認識されない → manifest.json書いて手動で拡張機能インストールしたら解決した話 https://qiita.com/beru_ike/items/b8c73be63f5f111008de 22. 国の公開資料 76 本から建設費 42 万件を作った。神奈川の塗装工の労務単価は 13 年で 1.6 倍 https://qiita.com/Mugaway/items/163c4239dd7b85369778 ■ Aws 1. AWS DevOps Agent Skills を作成するためのベストプラクティス https://aws.amazon.com/jp/blogs/news/best-practices-for-writing-aws-devops-agent-skills/ #TechFeed #開発 【PR】CursorのProプラン（月$20）が高くて迷っていた方に朗報です！ 下記の招待リンクから新規登録すると、初月が【半額（実質$10）】で使えます。 通常の無料枠だと物足りない方、1ヶ月だけ格安でフル機能を試してみませんか？👇 https://cursor.com/referral?code=WRLI1PLRTOLP

People still run five Claude Code chats on one repo Make each chat ask before it edits a shared file Plain English -> Claude Code builds the mod File touched by another chat in 30 min -> it stops Proceed, cancel or move to a git worktree -> you pick One silent overwrite used to eat an evening of diffs Now it costs one question and a click Each chat gets a worktree, not your main branch

Claude Code + Codex in one terminal, and they talk to each other 🤝 One prompt: Claude codes, Codex reviews, finds a bug Claude missed, Claude fixes it, Codex verifies. https://github.com/saiharsha03/codex-panel @claudeai @ClaudeDevs @bcherny @trq212 #ClaudeCode #claudemods #buildinpublic

📷 A Claude Mod for Better Task Management link in bio

tried claude mod and it quite fun

got all worried about a mod i was excited for possibly being vibecoded and saw the commits list was utterly devoid of claude

🚨 Claude Code just got a lot more customizable. Anthropic's new “mods” let developers change how Claude Code behaves. They can: • Rewrite prompts • Change the interface • Modify tool calls • Replace built-in features You can even ask Claude to write a mod for itself. One important warning: mods aren't sandboxed, so only install ones you trust.

This 𝗗𝗘𝗦𝗜𝗚𝗡.𝗺𝗱 stops Claude Code and Codex from 𝘀𝗵𝗶𝗽𝗽𝗶𝗻𝗴 𝗴𝗲𝗻𝗲𝗿𝗶𝗰 𝗔𝗜 𝗨𝗜. It's Google's open format. Stitch reads it natively. Claude Code and Codex read it once you add one line to AGENTS .md: "Read DESIGN .md before any UI work." Tokens on top, rules below: → Overview → Colors → Typography → Layout → Elevation & Depth → Shapes → Components → Do's and Don'ts The agent stops inventing a new color and spacing value on each screen, so 𝗲𝘃𝗲𝗿𝘆 𝘀𝗰𝗿𝗲𝗲𝗻 𝗹𝗼𝗼𝗸𝘀 𝗹𝗶𝗸𝗲 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗽𝗿𝗼𝗱𝘂𝗰𝘁. Claude Code users: also install the official frontend plugin: /plugin install frontend-design@claude-plugins-official Check your file against the spec: npx @google/design.md lint DESIGN .md Copy and paste it into your DESIGN .md 👇

Claude Code now supports mods, small plugins that change how it looks and behaves. They shipped in version 2.1.287. A mod is written in JavaScript or TypeScript and runs inside Claude Code. It can draw panes and buttons, rewrite prompts and tool calls, hold a risky command for confirmation, or add your own commands. Anthropic built features like /diff and AGENTS.md support as mods. Mods are on by default in the CLI and the desktop app's Code tab. You can ask Claude to write one for you, or install one with /plugin install. Anthropic's docs say mods are not sandboxed and run with your permissions, so install them only from authors you trust. claude plugin validate lists what a mod does before you install it. @claudeai @AnthropicAI #AI #ClaudeCode #Anthropic

with mods you can change the memory section of the system prompt in claude code

Countless useless Claude mods on my timeline. Same deal with custom prompts and skills, most will end up wasting tokens trying things, instead of doing actual, productive work.

Good Claude code mods tutorial /jev-memory mod is pretty cool And very happy seeing @treg_ai used here :)

Comme quoi le modding avec Claude ne permet pas que du slop, il a réussi à transposer les textes FR de DQV DS dans la version mobile ! Actuellement il finalise le projet et ça devrait être fini d'ici demain (manque de jetons) ! #DragonQuest