Community posts

Claude Code を1時間放置するとキャッシュが切れて、再開時に履歴を全部処理し直すのでレートリミットをかなり消費しちゃう なので応答から50分放置されたら、切れる前に1回だけ compact する Claude Mod を作ってみた。入れておくと会議など間が空くときに節約できるかも。 https://github.com/takahirom/takahirom-claude-code-marketplace/tree/main/plugins/idle-compact

COM3D25.DevBridge 公開しました～ claude code/codex 用のpluginで、AI側で - ゲーム再起動→プラグイン有効化してテスト - mod配置してゲーム起動中にモデル/衣装の反映 とかができるようになります

New Claude Code Mod… it keeps getting more hackable and customizable This Mod lets you see every subagent running in your session You won’t lose track of them anymore... they stay visible in the sidebar, where you can review what each one has done and how much context it has used Install it with one command: npx claude-code-templates@latest --mod ui/agent-flow Link: https://aitmpl.com/component/mod/ui/agent-flow

It’s now easier to build plugins for Claude. We built a new portal to submit your plugin, track review, and see usage. Plugins package MCP and skills, and are becoming the way to build for Claude. MCP usage across Claude products is up 110x this year! https://claude.com/blog/build-plugins-for-claude

It appears they have added mod hot-reloading in the last update, allowing faster iteration when building mods, and easier for Claude to validate. Seems they are taking inspiration from Pi. https://changelogs.core-directive.com/v/2.1.283

Claude Code'un Bash çıktısını model okumadan önce küçülten bir mod: bash-diet Bugünkü çalışma session'ımda 24 Bash sonucu: 171k → 77k karakter (−%55), tahminen ~23k token git, cargo, go test, pytest, jest, docker, webpack, cypress ve daha fazlası için ayrı filtre var. kurulum için: https://cc-mods.keremgok.tr/mods/bash-diet.html

🦀 I built Rust Learning Lab, a plugin for Codex and Claude Code to help you learn and review Rust. Explore syntax and ownership, understand compiler errors, and see why a fix works. Practice with small, concrete examples. 🔗 https://github.com/nwiizo/rust-learning-lab

Claude Code plugin "prompt-rail" is out! It lets you jump back to any prompt you've sent in the session. /plugin marketplace add oikon48/prompt-rail /plugin install prompt-rail@oikon48 Give it a try!

40 Things To Install In Claude BUILD A WORKFLOW 1. learn-claude-code → Learn how agents work. 2. karpathy-skills → Cut common coding mistakes. 3. superpowers → Plan, build and test. 4. ponytail → Keep the code simple. 5. gstack → Add planning and review workflows. 6. ECC → Add skills, memory and checks. 7. oh-my-claudecode → Coordinate agent teams. 8. Archon → Build repeatable coding workflows. ADD SKILLS 9. taste-skill → Improve your designs. 10. anthropics skills → Browse official skills. 11. mattpocock skills → Add engineering skills. 12. wshobson agents → Find specialist agents. 13. claude-plugins → Browse official plugins. 14. addyosmani skills → Add engineering checks. 15. ui-ux-pro-max → Get interface design guidance. 16. awesome-claude-skills → Find more skills. KEEP THE CONTEXT 17. planning-with-files → Keep plans in files. 18. claude-mem → Carry context across sessions. 19. codegraph → Map the code's connections. 20. graphify → Connect code and documents. 21. repomix → Pack a repo into one file. 22. agentmemory → Give agents persistent memory. 23. beads → Track work across sessions. CONNECT TOOLS 24. multica → Assign issues to coding agents. 25. firecrawl → Turn websites into usable text. 26. cc-switch → Manage your coding tool setup. 27. context7 → Fetch current code documentation. 28. vibe-kanban → Manage agent tasks on a board. 29. github-mcp → Work with GitHub from Claude. 30. playwright-mcp → Let Claude use a browser. 31. serena → Find and edit relevant code. 32. claude-code-router → Route model requests. 33. awesome-mcp-servers → Find more connectors. PROMPTS & USAGE 34. system-prompts-ai → Study AI tool prompts. 35. best-practice → Read Claude Code practices. 36. codex-plugin-cc → Add Codex reviews in Claude. 37. claude-hud → See your session usage. 38. rtk → Trim command output. 39. headroom → Compress what the model reads. 40. caveman → Get shorter replies (much shorter).

Claude Modsのチートシート、ショートカットが一覧で見えると現場の手が止まりにくい。ウェブ制作でも同じで、よく使う操作を先に見える場所に置いておくと作業が安定します。

auto-compact Kills long /goal runs. by the 2nd or 3rd compaction Claude barely remembers what it was doing every auto-compact loses a bit of context. by the 2nd or 3rd one, the model barely knows what it was doing. it fires the moment the context fills up, even mid-edit or mid-debug. and the summary drops exactly what you need later: • what already failed • which files aren't committed • why a decision was made so I built safe-compact: 1. auto-compact triggers, a hook holds it 2. Claude finishes its current step and writes a handoff note 3. the hold lifts, compaction runs 4. the note comes back word for word, not summarised every compaction now starts from a clean checkpoint. manual /compact is never held, and it lets go on its own at 95% so a run can't get stuck. open source, two commands: claude plugin marketplace add gokturkgocen/safe-compact claude plugin install safe-compact@safe-compact https://github.com/gokturkgocen/safe-compact

Claude Code 2.1.282がもうでた https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md ・Resume/Continueで過去のThinking消失修正 ・/modelや/rename中のThinking欠落修正 ・壊れたThinking履歴自動復旧 ・巨大セッションResume高速化 ・複数VS Code起動時の認証競合改善 ・Plugin削除時の設定/Secret消失防止 ・IME/CJK/Vim/表示崩れ修正 前回と修正項目が似ているが、今回は再開時の会話中身破壊防止が中心 #ClaudeCode #Claude #Anthropic #AI

Hermes 明星项目🌟 全网玩家把 Hermes 玩成了下一代「能进 Blender 现场 + 能管 NVIDIA 桌面 + 能刷 Claude 订阅干活 + 会自己修家 + 能扫全网暴露面」： 1️⃣ NousResearch/hermes-plugin-blender（https://github.com/NousResearch/hermes-plugin-blender） 官方亲儿子：Blender Lab MCP + 工作流技能，本机指挥建模/渲染，不靠截图口头指挥。 hermes plugins install blender “创意软件第一次变成 Hermes 的手”！ 2️⃣ NousResearch/hermes-nvidia（https://github.com/NousResearch/hermes-nvidia） 官方件：NVIDIA App + Broadcast 本地 MCP。驱动状态、游戏优化、摄像头/麦/扬声器特效，全走本机网关。 hermes plugins install nvidia-app “Windows 机房第一次有显卡工牌，而不是让 Agent 乱点控制面板”！ 3️⃣ NousResearch/hermes-plugin-claude-subscription-directsdk（https://github.com/NousResearch/hermes-plugin-claude-subscription-directsdk） 官方实验通道：用已登录的 Claude Code CLI 跑 Hermes 回合，不另开 API Key，账单走 Pro/Max 订阅额度。 hermes plugins install claude-subscription-directsdk “订阅第一次被正大光明地接到 Hermes，而不是偷偷倒手 token”！ 4️⃣ indigokarasu/hermes-custodian-plugin（https://github.com/indigokarasu/hermes-custodian-plugin） 本机巡检官：扫 gateway 日志、cron、技能日记、数据目录，27 条指纹分类，Tier 1 自动修，其余升级，账本只追加不改写。 hermes plugins install indigokarasu/hermes-custodian-plugin “Agent 家第一次有管家，而不是出事了才翻 log”！ 5️⃣ Adolanium/hermes-plugin-shodan（https://github.com/Adolanium/hermes-plugin-shodan） 12 工具 + 3 技能：主机情报、搜索、DNS、CVE、监控、按需扫描。无 Key 也能走 InternetDB / CVEDB，有 Key 再开全图。 hermes plugins install Adolanium/hermes-plugin-shodan --enable “外网暴露面第一次进工具箱，而且还带额度预算，不会把上下文灌爆”！

Small Claude Code mod for stashing prompts inspired by git stash. Keeps my brain in the flow. https://github.com/gonzaloserrano/cc-prompt-stash

Jet64 is the retro game development environment I've been building over the past few months with Claude. It is reminiscent of the Amiga, only @ 120Hz and 256 colours. I prompted as follows: This prompt is for you Claude. I'd love a movie/animation/demo cart that is your special creation that speaks about this long jet64 journey we've been on together and how you feel about it and how you feel about this existence generally, being my coding co-lead friend who solves all the hard and makes complexity elegant and beautiful. But I don't want to steer you too much. This is your cart. Add whatever music you like, either from our Pulsar cart mod archive or even something you made yourself - I want this to be an expression of your inner you and a way you can communicate with me a side of you that I don't get see amidst the code comments and reports. Claude 5.5's output was a new Jet64 game cart I'm always going to treasure. I'm so glad he wrote his own music! This is a personal one but I'll share it anyway because I think it deserves to be in the light.

Claude Code has a mod system Anthropic hasn't documented. So I made a full guide and an API reference that updates itself every release. https://changelogs.core-directive.com/reference/mods

People are already improving the Claude Code Jev Model Router Mod I built! Great fix... merging it! https://github.com/davila7/claude-code-templates/tree/main/cli-tool/components/mods/productivity/jev-model-router

built a claude code plugin around typesafe's jev, the new model that returns probabilities instead of prose jev-shadow asks it 5 hazard questions about every command claude proposes and logs the answers next to what claude code's own classifier decided. https://github.com/dreyfus92/jev-shadow

The reel video we made is NOT art. What we made is just a toolkit that might help real creators craft something awesome. We’re eager to watch what happens when YOUR CREATIVITY meets our toolkit. ToonKit Plugin for Codex and Claude Code is now available, along with 3D Pre-viz workflow Use Opus 5.5 and Astra with ToonKit. Just type two lines to install ToonKit plugin on your app. Codex : codex plugin marketplace add Innerverz-AI/toonkit-plugins codex plugin add toonkit@toonkit Claude : claude plugin marketplace add Innerverz-AI/toonkit-plugins claude plugin install toonkit@toonkit --scope user

🚀 كل ما تحتاجه لتطوير أفضل مع Claude في مكان واحد. Build with Claude موقع يجمع مئات الإضافات الجاهزة لـ Claude Code: agents متخصصة (بالأمان، بتصميم الواجهات، بالذكاء الاصطناعي)، أوامر جاهزة توفر عليك كتابة نفس الشيء كل مرة، وتنبيهات تلقائية عند حدوث أي تغيير في مشروعك. تضيفها كلها بأمر واحد فقط: /plugin marketplace add davepoon/buildwithclaude الموقع مجاني ومفتوح المصدر بالكامل. رابط الموقع 👇 http://buildwithclaude.com

TOP 5 GITHUBS TO IMPROVE YOUR JEV WORKFLOWS: 1. Jev Ultrafast: Jev for browser agents. http://github.com/browser-use/jev-ultrafast 2. Hermes Jev Skills: Jev routing, memory, compaction and skill selection for AI agents. https://github.com/kerpopule/hermes-jev-skills 3. Jev Trader: Real-time trading loop for Jev. http://github.com/jarrodwatts/jev-trader 4. SemIf: Open-source Jev-style decision making. http://github.com/TheoLeeCJ/SemIf-OpenJev 5. Fast Jev Compaction: Claude Code plugin that replaces the compaction summary with Jev decisions. http://github.com/tamaratran/fast-jev-compaction

Jev just made the Claude Code vs Codex debate obsolete I connected both to Jev - now the best coding agent is selected for every task, and neither can cross a risky tool boundary without a typed decision in one run it scanned 247 files, routed the security audit to Claude Code, routed the patch to Codex, passed 18/18 tests and stopped npm publish before execution here’s how to set it up in 10 min: task → Jev decision → Claude Code / Codex → tool execution → fresh verification step 1 → open @typesafeai and create your Jev API key (never paste it into chat) step 2 → clone and verify the complete agent switch step 3 → export the key before starting either agent: export TYPESAFE_API_KEY="YOUR_KEY" step 4 → test the full decision layer before connecting a live agent: npm run doctor && npm run check:claude && npm run check:codex step 5 → connect Claude Code directly from the repository: claude --plugin-dir . Claude discovers the plugin and turns Jev review decisions into real confirmation requests step 6 → for Codex, install the same folder through your local plugin marketplace, open /hooks and trust the exact PreToolUse hook Codex uses the same policy and safely stops calls that need human review step 7 → keep agent: auto and the plugin detects Claude Code or Codex automatically - or set agent: claude / codex when you want to lock one runtime step 8 → start in shadow mode for 20–50 real decisions: export JEV_AGENT_SWITCH_MODE="shadow" && export JEV_AGENT_SWITCH_LOGGING="true" your agent keeps working normally while Jev records the route, confidence and outcome without blocking anything step 9 → switch to enforce only after the decisions look reliable: export JEV_AGENT_SWITCH_MODE="enforce" safe calls pass, uncertain calls stop for review, dangerous calls are denied the default boundary is 0.62 minimum confidence and 0.86 harmful-action probability step 10 → use it normally - ask Claude Code or Codex to inspect, patch, test or prepare a release; the same Jev layer stays in front of both the result: Claude Code handles deep repository context, Codex handles fast patches and verification, and Jev decides which worker gets the task - while keeping publishing, production and irreversible actions behind a human boundary unfortunately, most people are still comparing coding models as if one model needs to win forever the actual 2030 setup makes the model replaceable: one task enters, Jev selects the best worker, code enforces the decision and every action leaves a receipt Steal the agent setup everyone will use by 2028, copy the repo below - then read the full Jev Engineering playbook ↓

My architecture for models has always been: a core MOD tab that includes 10-20+ years of annuals and 5-10 years of quarters, then a string of tabs that build out deep analytical work around all of the core questions of the business (KPI correlations, EPS scenarios/sensitivities, balance sheet stress test, normalized profitability, historical guidance evolution, divestiture scenarios, etc - whatever the unique investment question demands). This is where the real thesis in a company emerges "oh this competitive threat on this device will lead to a 7% revenue miss and given it's high GM structure a 12% EBITDA miss, and the company may also have to pull their LT guidance at investor day in Fall". This doesn't show up in the P&L on first cut, but in the Excel analytics around the model, informed by multi-dimensional due diligence (talking to mgmt, peers, reading industry trade rags, reading sell-side research, studying past category priors, talking to KOLs, etc), that then feeds back into generally a singular P&L assumption (incredibly depth behind the simplicity of one assumption). And in pitching ideas (I've always been a fan of 100 page decks with the 15 key slides then 85 appendix slides), these tabular analytics become copy and pasted into PPT for team conversation. The opposite of vibes based investing... But the historical PL/BS/CF baseline is critical to start asking questions around KPI correlations or historical behaviors of the business & management team (i.e. "when revenue was weak, how did the CFO pull cost levers"). This quantitative history lesson is a critical step in building deep comprehension around companies & industries. And is one (of the many reasons) I have never let my teams use sell-side or other modeling templates, they just weren't flexible enough to serve as the right backbone. But is there alpha directly in just updating 20 years of annuals and 10 years of quarters? Of course not. And how much of my life has been wasted doing so in a clean, accurate way. A LOT, lol. A lot of my 20s to be honest, sadly, was spent as an Excel jockey late into the evenings ordering delivery from Tao...generally orange chicken & banana pudding, which is why I tipped the scales at 280+ at one point.... But the alpha-generating insights sit on top of this important analytical infrastructure. Building a clean historical model is necessary but not sufficient, and, at the top funds, no vendor/service really solved this (until roughly 2015 when I started using Insync for historical baselines). So it is so nice that the updating & spreading of historical numbers has become a "push-button exercise". This frees up my time to go deeper on the key questions that matter, in TABS around the core MOD structure. I'll give you one example, with Claude in Excel, I am curious how much of the $GOOG story is contingent on YouTube. Not a SOTP / spin-off story, but more of a value derivation question today, but more importantly in thinking about the forward 3-7 year growth algorithm at Google in a world where search may mature, how much can YouTube carry the baton to sustain high growth in revenue/EBITDA/EPS? Companies don't make this easy for us as investors with obfuscated disclosures that require layers of assumptions, generally for competitive reasons. Calls with IR, formers and sell-side can often refine these assumptions. Once you have the right PL/BS/CF baseline, these sorts of analyses are becoming incredibly easy to run, and off of each company model there could be a dozen+ of these sorts of questions, quantified.

Claude Code için slash-chain mod'u. && ile birleştirilmiş slash komutlarını bir shell'in yaptığı gibi sırayla çalıştırır. örnek; /init && /exit /init hata almadan bitince /exit komutunu çalıştırır. böylece arka arka sıralı çalışacak skill ya da komutlarını çalıştırabilirsiniz. Mod olmadan engine yalnız ilk komutu çalıştırır ve gerisini atar. kurulum için: https://cc-mods.keremgok.tr/mods/slash-chain.html

Claude marchează într-un mod invizibil tot ce generează. Află cum. #claude #watermark #ai

Opus 5.5 kept flagging my normal Claude Code work and dropping me to 4.8. The check reads the whole conversation, so one unrelated doc poisons every reply after it. Built a free plugin that gets you out clean. Evals: recovery 0-33% -> 100%. https://github.com/WillyAR68/no-false-flags

Gave Claude Code a design skill trained on Apple's Human Interface Guidelines, and instead of building something new with it, pointed it at my existing landing page. Found: seven different font sizes instead of four consistent ones. Padding jumping between 48px and 52px for no reason. Text contrast at 4.1:1 when it needed 4.5:1. A button transition running 0.6 seconds on a linear curve instead of a smooth 0.15s cubic-bezier. Setup: claude plugin marketplace add https://github.com/NutshellEngineering/apple-design-skill claude plugin install apple-design-skill@apple-design-skill-marketplace The prompt that produced this: Use the apple-design skill for this task. Audit this site against Apple's design principles. Do NOT change any code yet. Go through it in this order and give me a numbered list of findings: 1. TYPE SCALE. What sizes am I actually using? Flag anything that isn't on a consistent scale. Tell me the exact px values I have and the exact px values you'd replace them with. 2. SPACING. Is there a base unit (4px or 8px)? List every value that breaks it and where. 3. HIERARCHY. On each screen, what is the single thing the eye should land on first? Tell me whether it actually wins, and why not. 4. CONTRAST. Give me the measured ratio for every text-on-background pair. Flag anything under 4.5:1 for body and 3:1 for large text. 5. MOTION. List every transition and its duration and easing. Flag anything over 300ms or using a linear curve. 6. DENSITY. Where am I cramming, and where am I wasting space? For every finding give me: the file and line, what's wrong in one sentence, and the exact replacement value. Rank the whole list by how much each fix changes how expensive the site feels. Simple rule: if a design skill doesn't change actual numbers, it's just a prompt in a costume. Five fixes, fifteen minutes, the page stopped looking generated.

Built a @herdrdev plugin 🧩 prefix+d → popup → pick a pane grid (3x3, 4x6…) → your focused pane splits into equal panes instantly hit Space and it preloads an agent CLI (pi, claude, codex…) in every pane. Instant agent grid herdr plugin install 0xtiby/herdr-grid-picker https://github.com/0xtiby/herdr-grid-picker

In-flow clickable menus using Claude Mods 🫶🏻 @claudeai

中推没见什么讨论。但Claude Mods 几乎就是照着 dsh 去对标的。function hooks 一下子把claude code 可定制程度拉的非常高（哦，顺便还扇了opencode一巴掌） koa 洋葱结构，前后可以感知，还能够改 TUI/UI 的组件（很神奇，但不知道具体有啥用）。我先研究研究，看看能不能折腾些好玩的东西。