Community posts

Biggest story today: an OpenAI safety researcher quit publicly and said the risk culture is broken. Worth tracking if you build on frontier models. https://techcrunch.com/2026/10/03/openai-safety-employee-resigns-claiming-the-companys-culture-is-broken/ The tools: Claude Code v2.1.288 lets mods read your highlighted text, and cloud sessions get a built-in gh api. https://github.com/anthropics/claude-code/releases/tag/v2.1.288 ponytail pushes your agent to delete code, not write more. ~1,900 stars in a day. https://x.com/trending_repos/status/2106718156463182216 Ten open-source projects for giving agents real tools without real keys. https://x.com/Promixyz/status/2106693751066264008 ten by @e2b replica-skill clones an app, then fixes what its users complained about. https://github.com/Jakeschincariol/replica-skill live-panel-skill turns a JSON config into a moving architecture diagram. https://github.com/ythx-101/live-panel-skill live-panel-skill by @YuLin807 Agents are getting real access. The tooling around them is finally catching up. Full episode: https://askconsulting.io/podcast/api/episode?d=2026-10-04&e=2026-10-04

天の声 Claudeの自作Mods。プロンプトやAIの出力に対して天の声(つっこみ)が入る。役に立つかもしれないし立たないかもしれない。 元ネタは古のパソコンゲームの天の声システム。

Claude Code can now modify Claude Code while it’s running. Anthropic’s new Mods are TypeScript functions that can: rewrite prompts block or rewrite tool calls approve or deny permission requests redact secrets change UI replace built-in features and add entirely new functionality. this is where “prompt engineering” starts looking increasingly small. the interesting unit is becoming: behavior + tools + permissions + interfaces. not clever paragraphs in a textbox. also worth noting: Anthropic says Mods aren’t sandboxed. power cuts both ways.

I didn't want to type /cost every few minutes in Claude Code, so I made a small mod that shows it under the prompt. On Pro/Max it shows how much of your 5h and weekly limits you've used and when they reset. On the API, dollars. Tokens per turn too. #claude #ai

天使と悪魔 Claudeの自作Mods。漫画でよくある、天使と悪魔が脳内で対立するやつ。プロンプトに対して議論を始める。議論を続けるを選ぶと１ターンずつ進める。

@ClaudeDevs I've created 11 different useful and fun mods, they're open source and available at https://github.com/theonly1me/claude-code-mods I'll create more mods that I find useful. Here's a quick demo of them, install instructions in the README

@ClaudeDevs We benchmarked the idea behind "You should know" against what Teknesyum Core already does: a closing note the main agent writes itself. Closing note: 12/12 buried decisions caught, zero extra calls. Side observer: 8/8, one extra call per long turn. https://github.com/Teknesyum/Teknesyum-Core

Can somebody mod the antigravity app to use with custom api and models. That shit is really good at doing stuff just the nodel limits are abysmal. Only if we could use claude or gpt models with it from our subs.

claude code has mods now. a few lines of typescript that change how it behaves, redo the ui, or swap in your own features. or you just ask claude to write the mod for you. they install with /plugin and work in the desktop app too. that's the part i care about, i use the app not the terminal. anthropic also shipped one called "you should know", a side agent that watches claude's output and flags things you might miss. what's the first thing you'd mod?

Claude Code just got mods, and I think it's a bigger deal than it sounds. Small TypeScript functions that hook into its events: block risky tool calls, redact secrets before Claude sees them, require a confirm before anything touches prod. Not sandboxed, so trusted only.

made a Claude Code mod that plays TikTok in a pane beside the transcript while Claude works /tiktok to open. j next, l like, p pause.

Claude Code just became programmable. Most people scrolled right past it. On Oct 1 Anthropic shipped "mods": tiny TypeScript functions that sit inside the agent loop and change what Claude Code actually does. Not settings. Not CLAUDE.md. Real code that runs on every event. what a mod can do: > REWRITE - edit your prompt before it ever reaches the model > BLOCK - stop a tool call cold. rm -rf, force pushes, prod deploys > REDACT - strip API keys and secrets before they leave your machine > DECIDE - answer permission prompts with your own rules > REDRAW - replace parts of the interface itself the wild part: Anthropic already rebuilt its own diff pane and AGENTS.md loader as mods. this isn't a toy feature. it's how Claude Code is being built now. and you don't even have to write them. describe the mod you want and Claude Code writes it for you. the catch almost nobody mentions: mods are NOT sandboxed. they run with the same access to your machine as Claude Code itself. installing a random mod from GitHub = running a random program on your laptop. read it before you run it. needs Claude Code v2.1.287 or later. your first mod, paste this into Claude Code: "Write me a mod that blocks any tool call containing rm -rf, git push --force or DROP TABLE, and tells me why it was blocked." watch every event go through the stack 👇

A file tree with git status and a preview pane, right inside the terminal. Tried Claude Code mods today. I built one tiny feature, the one I've missed most: viewing files right inside the terminal, no more switching to the editor just to peek at a file

anthropic 给 claude code 上了个 mods 说白了就是官方下场做行为定制 以后调教 agent 不用再靠玄学 prompt 了 这波有意思的点在于 coding agent 都在从拼模型转向拼工作流 谁把开发者的习惯吃透，谁就赢 等等党可以先看看社区 mods 生态能不能起来

Cat, coding, got some football on while I work. Best Sunday setup. Got 4 Claude tabs open and yes that is space invaders in a Claude Mod and I will post more about that once it’s finished

[92/365] I thought it would be a semi day off but here I am at 2am writing this ahah Today I focused on my Dev workspace and how I can use Hermes to produce better code The idea is simple 1 - I create a task in my kanban 2 - When I'm done I move it into Review 3 - Hermes receive a notification, clone the project, create a workspace with the correct branch 4 - Hermes run a code review workflow 5 - Hermes run a code security check 6 - Hermes run a test check using e2e 7 - Hermes move the task into Human validation 8 - I merge the PR I'm also building a game that can be integrated into as a Claude Code mod to make indiehacker connect while they code

animal crossing brewster owl mod . how to make brewster an owl in animal crossing . computer make brewster an owl . brewster pigeon to owl tf . claude vibecode me animal crossing but brewster is an owl .

this mod would be cool af in codex in claude is completely useless since opus 5.5 dropped i cant even finish the limits of 200$ plan

I built the perfect mod for pacing your Claude usage

@FracSlap Claude Code Mods are fun. Sometimes you just have to enjoy the little things

ClaudeCodeのMod で動くドット絵ペット作ってみました！ https://github.com/i-noma-ru/claude-dotpet

Built a Claude Code mod that keeps my prompt cache warm through the day. Pings chats 5 min before the 1h cache expires and shows a countdown in the status line. Does it through work hours and a limited number of pings. Ran it on a week of my logs, and through all of my cold rebuilds and approximations of my weekly usage against API prices, the mod saves... ~1.3% lol Anyone else experimenting with this though?

Morning Tech Digest | October 4, 2026. Coverage through 6:00 a.m. Pacific, with dated catch-ups. 00:00 Sol ultrafast clarification 08:28 Tibo’s simpler-tools priorities 12:54 Inbox zero and validation questions 18:36 Hermes Gadget community SDK 25:59 Claude voice-data choice 32:31 Gemini access changes (catch-up) 36:45 Harness Mods bridge (catch-up) 41:48 Cloud spending caps (catch-up) 47:55 AI-connected yeast research (catch-up)

New video 🎬 Claude Code can now rewrite itself mid-conversation. Mods can block commands, reroute calls, add live panels, and see every API key you have. What that really means 👇 https://youtu.be/46sODzFRwEU

I built a Claude Code plugin that turns a codebase into a narrated video. It reads the repo, explains the architecture and code structure, checks its claims against the source, then renders the walkthrough with code, diagrams and narration. https://github.com/OnurGumus/codebase-video

What claude code mods do y'all want? Not some fluffy beatification mod. Something which will solve you actual problems?

Another cool thing about the Fourier Transform (and its inverse transform) is how it repeats itself mod 4 when being applied over and over. I was reading this book and one section was going over Fourier Duality Properties so I drew some pictures by hand and then asked Claude to make these pictures. I remember people trying to memorize duality properties back in undergrad but this cycle graph I made I feel makes it easier. Sharing in case it’s useful to others.

7/ segmem 给 Claude Code 加长期记忆。 Identity、Procedural、Episodic Memory 存在本地 SQLite，跨 Session Recall。 https://claudemods.ai/mods/segmem

Claude Code 的 Mods 上线才 3 天，我把社区翻了个底朝天。 筛了几十个项目，最后留下 10 个我认为最值得装的。 直接上名单。🧵👇

You should know that you should install the 'you-should-know' plugin in Claude Code *TODAY*. It surfaces important information from complex tasks that you might miss. Another example of the evolving "Attention Interface". Turn it on: /plugin enable