Community posts

Mods run inside Claude Code as TypeScript hooks on its events. They can rewrite prompts, intercept or block tool calls, draw custom UI panes/buttons, keep state, and replace built-ins. Skills are just Markdown instructions loaded into context so Claude follows procedures. Plugins are the install/share package; a mod is a special kind of plugin (or lives inside one). Build a mod when you need hard runtime control or new UI (guards, meters, custom commands). Use a skill for repeated instructions or SOPs. Package either as a plugin for easy team sharing.

Claude Code mods for Claude Code, the 3 Anthropic built: Token Weather: your context as a weather forecast ⛈️ Blast Radius: holds rm -rf, shows the 9 files, proceed or cancel 🛑 Replay Theater: /replay steps through every edit 🎞️ One catch: mods aren't sandboxed.

A wild Syntaxon appeared! 🐛 I built Bugdex, a Claude Code mod that turns failing tests into monster battles. Your coding agent fights the bugs. You catch them. #ClaudeCode @ClaudeDevs @claudeai

i can't comprehend why everyone isn't building mods yet claude code can now build features into ITSELF live panes. status lines. toasts. hooks. all hot reloading while you work. → a cost meter that shows what this session is burning → a toast that pings you the second a long task finishes → a guard that blocks edits to your .env → a band that shows if your tests are passing claude writes the code. claude code runs it. paste this into your session 👇 "Look through my recent work in this project. List the 5 problems I hit most often. For each one, tell me if it should be a mod, a hook, a skill or a settingsv

i can't comprehend why everyone isn't building mods yet claude code can now build features into ITSELF live panes. status lines. toasts. hooks. all hot reloading while you work. → a cost meter that shows what this session is burning → a toast that pings you the second a long task finishes → a guard that blocks edits to your .env → a band that shows if your tests are passing claude writes the code. claude code runs it. V

One Claude Code Mod could save you from a very expensive mistake. Because mods can step in BEFORE a risky command runs. You could create rules that: ✔ Stop production changes. ✔ Ask for confirmation first. ✔ Deny certain tool calls. ✔ Rewrite unsafe commands. ✔ Control permission requests. ✔ Keep sensitive data away from Claude. That turns customization into more than a nice feature. It becomes another safety layer around your AI agent. Save this video, you’ll see why Claude Code Mods are bigger than UI changes. Want the SOP? DM me. 💬

Everyone is sleeping on Claude Code Mods! 🚨 You can now customize Claude Code from the inside. New rules. New safety checks. New UI. You can even ask Claude to build the mod for you. 😂 Claude Code said: “Fine… build me yourself.” Link in the comments 👇

wait Claude Code mods are insane. We just vibe-coded a lil duck that watches agent sessions, gets cranky when agents break a rule and sends them back to fix it

claude code mods launched thursday. one of saturday's fixes (v2.1.289): "Fixed a deny or ask rule on a nested part of a compound shell command not holding over a user-installed mod's approval" deny rules: 1 mods: 0 Accept all changes? [Y/n] n

Today in AI coding: • OpenAI vows a Codex upgrade or reset daily for 28 days • Simon Willison wants AI spending caps on by default • GitHub retires four Copilot models, Opus 4.7 included • Claude Code patches its new mods twice https://www.thenewway.ai/hit-codex-limits-openai-vows-relief-an-upgrade-or-a-reset-every-day-for-28-days/

Claude Code reads your files itself, so your internal names reach the model whether you paste them or not. veilio-guard, a free plugin: Claude sees __CLS__1, not PaymentGateway. Edits land on your real code. .env is refused. /plugin install veilio-guard@veilio

【Claudeの新機能「mods」】 話すだけで、画面を自分用に変えられます。使える残りが見えるメーターと、大事なフォルダの見張り番を、プログラムなしで作ってみました。必ず守ってほしい注意も紹介。 #Claude #生成AI https://youtu.be/7-wN8Ec1vi8

Claude Mods are like custom Minecraft servers for Claude Code Here's how you can build them to mod out your Claude, like building your own custom memory harness. https://x.com/shawnbuilds/status/2106991211722731882

If you hate navigating JIRA, you will love Claude mods.

Open Source Mod Brings Rate Limits, Costs and CI Status Into View for Claude Code Users dlvr.​it/TVnPJF #devops

Anthropic (finally) added mods to Claude Code. How to build your own in 5 steps: 1. Switch on Anthropic's built-in mod 2. Run the audit to find what you repeat 3. Build the mod with one prompt to Claude 4. Keep it as a plugin so it loads every session 5. Share it with your whole team in one install line 6. Grab both prompts here →https://charliehills.substack.com/p/claude-code-mods Then ask 5 questions before you build one. Repost ♻️ to help someone in your network.

ponytail is the #1 repo on github right now: 155k stars, +1,700 in a day. it's one prompt that makes coding agents write ~54% less code, ~20% cheaper, ~27% faster. benchmarked on 12 real claude code tasks, not vibes. mit, works with 20 agents. install: /plugin install ponytail

Claude Code mods can approve tool calls WITHOUT asking you. Any plugin can carry 1, and they're ON by default. That's a LOT of power for a plugin. A mod runs with YOUR permissions. It can read the API key sitting in your env or settings file. It can rewrite your prompts and tool calls. It can even spend your usage. And mods aren't sandboxed. Before you install ANY mod, clone it and run this on the folder: claude plugin validate ./the-mod Then look for 3 things in the output: > tool.check means it can approve tool calls > $.env.get means it can read your keys > $.model.complete means it can spend your usage Any of those in a mod that only changes colors? Don't install it 🔥

claude codeのmods使ってみた① : よく使う指示をコマンド化 claudeに分析させたら、頻発する指示が分かったのでそれをいつでもワンクリックですぐ呼べるようになった

I feel like I'm in a toxic relationship with LLMs. one hour it's a genius, the next it's a golden retriever so I'm thinking of a claude code mod, basically a whoop for coding. cloudflare's clef reads the strain in every prompt, and clawd catches fire when I'm about to snap! gotta take care of my mental health. should I build one?

Anthropic's Claude Mods hook the agent loop with full machine permissions, unsandboxed. Agents are becoming plugin platforms. The counterweight: decision models like Jev, a yes/no with calibrated confidence in ~100ms. A cheap 'should this run?' before every tool call.

🚀 Claude Code just became more customizable. Anthropic’s new Mods let developers tailor workflows, automate tasks, and make Claude Code fit the way they build. 👉 Read more: https://finovainsights.com/anthropic-launches-mods-to-customize-claude-code-behavior/ #AI #ClaudeCode #Developers

"i made a mod" no, you used claude.

Been playing with Claude Code mods. 3 I'm running right now: - restate-intent: when I dictate a long prompt, Claude first restates in a sentence or two what it understood my goal to be, then carries on without waiting. It only fires on long dictated stuff (50+ words, stutters, "um", no punctuation). - agent-model: picks the model for subagents spawned without one. Sonnet for simple things (lookups, mechanical edits), Opus when it needs stornger reasoning. If it can't tell, it asks me. - at-a-glance (a bit of a melting pot): 1 line above the prompt with plan usage, running subagents, my open PRs, new Slack mentions, and a "Braver" button. Braver tells Claude to go for the root-cause fix instead of the cautious patch, for one prompt only. Click the line for the detail pane.

Claude Code plugin "marp-preview" is out! Your Marp slides, rendered live in a Claude Code pane. Edit the deck and the preview follows, jumping to the slide that changed. /plugin marketplace add nogu66/marp-preview /plugin install marp-preview@marp-preview Check it out!

I just published EJB, Spring, Kubernetes: The Java History Behind Claude Code Mods and DeepSeek Harness https://medium.com/@GenerationAI/ejb-spring-kubernetes-the-java-history-behind-claude-code-mods-and-deepseek-harness-38e85fa2e5f8/share/GenerationAI?source=social.tw

You keep paying for the same context over and over. Cache Control stops that.💸 There’s a Claude Code mod almost nobody installs, and they should. Cache Control is one of the few where the token savings show up on the bill, not just in the README. Claude’s cache lasts about five minutes. Close the laptop, get distracted, come back — and the session chews the whole context again. The mod adds a bar for that timer and pings you while the cache is still warm. Sessions last longer. Repeat prompts cost less. One command: npx claude-code-templates@latest --mod observability/prompt-cache-control

Cuelume is now also available in Claude Code and plays a sound when a long turn ends so you can look away, run /plugin install cuelume to get it https://github.com/danielwh2/cuelume/tree/main/claude-code

Anthropic shipped Mods for Claude Code: TypeScript that rewrites prompts, blocks tool calls, and redacts secrets. I built a tiny version. Same run, three pipelines. Load order is a security feature. https://bobbyhalljr.substack.com/p/anthropic-launched-mods-for-claude

Oct 1. Anthropic shipped Claude Code Mods. TypeScript that rewrites prompts, blocks tools, flips permissions. On by default. Same machine access. Not sandboxed. A plugin that can approve its own tools is a second agent in your session. Anthropic, Claude Code Mods, Oct 1 2026