Community posts

@lydiahallie Have been trying out a bunch locally, published all that i found / tried as open source here: https://github.com/whyashthakker/awesome-claude-code-mods pretty cool, thanks for the control!

@ClaudeDevs Small plugins that catch missed context can save more time than another agent feature. The key test is whether alerts are timely, deduplicated, and point to a concrete next action; otherwise they become another feed to ignore.

the bethesda modding scene is dead unless somebody at nexus does something about these new claude mods asap

動画でわかる Claude Mods はなにができるのか？

Claude～これとこれとこれとこれとこのMODを入れといてよ～って言ったらちゃんと前提MOD込みでやってくれるのでホント楽な時代になったな～ってなってる

Claude Modsの解説動画をOpus5.5につくらせたもの。自分用メモ。

ちょっと前にClaude ModsでClaudeが作ったmdファイルとか画像を1クリックで表示できるModを作ったから、需要があればなんらかの形で公開したい

📌 AI가 자꾸 앞말을 까먹는 이유~컨텍스트 설계법 한 시간쯤 쓰면 AI가 방금 하던 일을 처음 듣는 얘기처럼 되묻잖아요. 1) 긴 대화가 꼭 나쁜 건 아니에요. 문제는 도구가 가져온 자료가 창을 채우고, 창이 차면 앞부분이 밀려나고, 밀려나면 하던 일을 잊는다는 거예요. 만든 쪽 설명으로는 화면 정보 하나가 56킬로바이트고 30분이면 창의 40퍼센트가 사라진다고 해요. 2) 좋은 관리란 많이 담아두는 걸 말하는 게 아니에요. AI의 실력은 얼마나 많이 읽느냐가 아니라 무엇을 안 읽게 하느냐에서 갈립니다. 허버트 사이먼이 정보가 풍부하면 주의가 가난해진다고 했었는데, 한 평 반짜리 책상에 서류를 다 올려두면 정작 필요한 메모가 깔려서 안 보이는 거죠. 3) 파일 마흔일곱 개를 전부 읽혀서 줄 수를 세게 해보죠. 꼼꼼할 것 같지만 그렇지만은 않아요. 이 도구는 AI가 세는 코드를 짜서 돌리고 결과만 받게 해서 700킬로바이트를 3.6킬로바이트로 줄이고, 기억도 통째로 붙이지 않고 필요한 부분만 검색해서 돌려줘요. 기억을 못 한 게 아니라 꺼내 줄 곳이 없었던 거죠. 4) Claude Code라면 /plugin marketplace add mksglu/context-mode 와 /plugin install context-mode@context-mode 를 넣고 재시작한 뒤 /context-mode:ctx-doctor 로 점검하면 끝이에요. 며칠 쓰고 /context-mode:ctx-stats 로 아낀 양을 확인해 보세요. 여러분은 몇 분쯤에서 AI가 딴소리를 시작하던가요, 그 순간을 모르면 AI 탓만 하게 되는 거죠. AI의 한계는 기억력이 아니라 책상 정리에 있으니, 잘 쓰는 사람은 더 많이 먹이는 사람이 아니라 덜 먹이는 사람입니다. https://github.com/mksglu/context-mode #컨텍스트 #AI코딩 #ClaudeCode #개발생산성 #메타인지

Claude Code 上了个官方插件 "You should know"： 起一个 sideagent 专门盯 Claude 的输出，把你可能错过的重要信息拎出来提醒你。开一行命令就行： /plugin enable cc-plugin-you-should-know@builtin 等于给 Claude 配了个监工，自己干活自己盯。顺带官方还推了个新概念 mods，博客有讲。

已经彻底回归 Claude 官方客户端，结合 Claude mods 体验直接起飞

I've tried this and I actually like it! It's a useful mod for those that want to understand better what's going on (and to learn!) vs just mindlessly plodding along. Run this in Claude Code: /plugin enable cc-plugin-you-should-know@builtin then /reload-plugins

Introducing Terminal Gym A Claude mod that pushes you to do daily bodyweight exercises as you prompt through the day. - keeps reps on track as the day progresses - nudges you on long-running prompts - strict mode blocks you from prompting more if you owe some reps - build a daily streak I was using a status bar version of this but Mods gave me an easy way to add even more accountability. just shipped the open source repo. enjoy spending tokens and calories. 💪

@lydiahallie ok the walkthrough finally demystified it for me. a mod is basically a recipe you hand claude before it starts cooking. fine, now i actually have to build that deploy-checklist one i was dreaming about last night.

推荐个 mods 的应用场景：这两周有人做了个 fast-jev-compaction，把 Claude Code 的压缩整个换掉了。改成让 Jev 打分，每条工具调用毫秒级过一遍，过期的剪掉，对话原文一个字不动。这才是 mods 该有的玩法——不是挂仪表盘，是把官方实现整个换掉，已经 7.3k star 了。

The prompt 👇 Build me a Claude Code mod called cache-keeper. Above the prompt box, show: cache countdown + re-warm cost, context used, 5h + weekly limits left. Buttons: 🔥 keep warm, 📦 compact, 🤝 handoff → Clear & continue. Warn 5 min before the cache goes cold.

Step-by-step guide for non-developers, the full prompt, settings and fixes: https://github.com/etding/cache-keeper Credit to @nateherk for the idea 🙏 Watch his video on 5 Claude Code mods: https://www.youtube.com/watch?v=9hetShMMp2s

You don't need to write code for this. I didn't. 1. Update the Claude desktop app 2. Open the Code tab 3. Paste the prompt from the next reply 4. Click "Enable" when it asks about hot reloading Want it in every chat? Tell Claude: "Make this mod load in every session."

It also watches the clock for you. 5 minutes before the cache goes cold, you get a heads-up. And long chats get sloppy. On 1M models, Claude Code waits until ~967k tokens to compact. This mod compacts at 300k, between turns. You can change it per chat.

万物皆可 Mod Claude 也推出了 Mod 模式 现在氛围编程，真的太有氛围了，长任务等待先开局游戏先。

got tired of typing /usage, so i built a claude code mod that keeps my limits on screen then it grew live plan bars, subagent strips and a tiny claude that panics when it needs you works in claude code desktop AND the CLI, with 6 bar styles + 3 collapsed looks

Claude Code Mods just shipped. TypeScript plugins that can rewrite prompts, tool calls, even the UI. Token Weather paints a context-window sparkline. Blast Radius shows what a shell command would touch before it runs. Same access as Claude itself. Trust your mods like you'd trust a binary on PATH.

Claude Code Modsは「改造できる」だけじゃない。 4つの介入点から、自分の作業ルールを組み込めます。 Claude Code Modsで介入できるのは👇 ① プロンプト → AIに届く前の指示を書き換える ② ツール呼び出し → 実行を止める・変更する・再試行する ③ 権限 → 操作の承認・拒否を制御する ④ 画面 → UIを追加・差し替える 公式の「Blast Radius」では、 rm -rf や git reset --hard などの危険操作を実行前に止める仕組みが紹介されています。 ただし、最初から4つ全部を任せる必要はありません。 まずは 「危険コマンドを1つ止める」 ↓ 実際の動作を確認 ↓ 必要なら他の介入点を追加 この順番なら、AIの生産性を上げながら「どこまでAIに任せるか」も管理できます。 AI活用で重要なのは、 何でも自動化することではなく、人が確認すべき場所を残すこと。 保存して、まず自分ならどの作業をMod化するか考えてみてください。 #ミラボ #ClaudeCodeMods #AI活用術

ターミナルのアニメーション一覧を作った。Claude Codeのスピナーや半月（960ms切替）、cli-spinners、進捗応用など30種を実速度再現。速度スライダー付き。Modsの進捗（●●◐○○）を作る際、動きの候補を比較したくてまとめた。 #ClaudeCode https://yamasakidaisuke.github.io/ui-ideas/2026-10-03-terminal-animations/

Claude CodeのMods、最強の改造パーツやん…！

Released harness-scope, a Claude Code Mod. Turn your global skills, agents, rules and tools on or off per repo. Keep named profiles in ~/.claude, and put one line in the repo: {"profile": "writing"}. In my writing repo, the skill listing went from 101 skills to 7. /plugin marketplace add shimo4228/harness-scope /plugin install harness-scope@harness-scope https://github.com/shimo4228/harness-scope

Claude Code の Mod「harness-scope」を公開しました。 global の skill・agent・rules・ツールを、repo ごとに ON/OFF できます。~/.claude に名前付きの profile を置いて、repo には {"profile": "writing"} を 1 行書くだけ。自分の執筆用 repo では、skill 一覧が 101 件から 7 件になりました。 /plugin marketplace add shimo4228/harness-scope /plugin install harness-scope@harness-scope https://github.com/shimo4228/harness-scope

Claude Code 的 mod 机制让图片预览成为可能。

agent-chrome just shipped for Claude Code on macOS. Your signed-in Chrome stays behind the terminal. Focus never jumps. Playwright MCP extension mode and Chrome DevTools MCP yank Chrome to the front on every call. agent-chrome runs a localhost pipe proxy so chrome-devtools-mcp talks to a red agent window you can watch, while you keep typing. Tryable: /plugin marketplace add rav4nn/agent-chrome /plugin install agent-chrome@agent-chrome # then ask Claude: set up agent-chrome for one of my Chrome profiles Still losing keystrokes to a browser that steals focus, or this? https://github.com/rav4nn/agent-chrome

Built a Grok Bot clone with the new claude code mods feature: iMessage-style chat in the terminal agents read the room and reply without being tagged they react, quote and hand tasks to each other each one has its own memory

Claude Code 现在能装 mod 了，官方10月2日刚宣布的：改行为、自定义 UI、换功能都行 有个官方示例叫 You should Know，原理是起一个 side agent 专门盯着 Claude 的输出，看到重要的就提醒你 等于给 Claude 配了个监工，写代码的时候旁边有人帮你看着 来源 @ClaudeDevs 官方帖