今日必看

Build Plugins for Claude

Shopify's own code treats silence as consent on every store it hosts

My coding agent pushed a commit deleting every file on main

Show HN: CK3DNA – searchable CK3 character DNA and coat of arms codes

Code is the future and present of hardware engineering

Show HN: I couldn't deal with another Claude Code tab

Yes, Claude can do nine loops

He who does not code, neither shall he eat

AI agent hacks government website for first time: why this breach matters

I gave my AI agent one harmless permission. It became a backdoor for everyone

Foundry hosted agent isolation with Microsoft Agent Framework

Show HN: An open-source launch kit to help you vibe code your knowledge base

Show HN: Web search agent for finding leads

GitHub Copilot app for Beginners: How to build custom workflows with canvases

Most tools give you a fixed set of screens and ask you to fit your work into them. But what if you could start with the workflow you want instead and have the interface take shape around it? That’s the idea behind canvases in the GitHub Copilot app . A canvas, also called a canva

Agentic autofix now uses Copilot Memory

Agentic autofix now uses Copilot Memory for customers who’ve enabled it. When you use agentic autofix, it reviews existing memories for context that can help resolve security alerts. When it creates a fix, it stores the fix pattern as a memory for future use. These memories can h

Updates to GitHub Copilot for Slack and Microsoft Teams

GitHub Copilot in Slack and Microsoft Teams now gives you more context, more control, and a clearer path from conversation to GitHub work. Whether you’re sharing files in Slack or images and forwarded messages in Teams, Copilot can use more of the context already in your conversa

Default Enablement of Copilot features for Copilot Business and Enterprise

We’re introducing a new global default policy for generally available GitHub Copilot features and supported client capabilities in enterprise and organization Copilot settings. For the next 28 days, you can configure this policy, but it won’t affect feature access for your users

GitHub Security Lab 发布 LLM 驱动的 Fuzzing Taskflow，自动完成 C/C++ 项目模糊测试全流程

GitHub Security Lab 的 Antonio Morales 开源了基于 Taskflow Agent 的 Fuzzing Taskflow，指向 GitHub 仓库即可自动识别入口点、编写 harness、运行 AFL++、读取覆盖报告并分诊崩溃。

Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning

arXiv:2609.28963v1 Announce Type: new Abstract: Group-based reinforcement learning (RL) methods, such as GRPO and its variants, have become a leading paradigm for training reasoning and agentic large language models (LLMs). While their group-normalized advantage estimation is rel

GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI

arXiv:2609.30147v1 Announce Type: new Abstract: Large Language Models (LLMs) typically exhibit a performance profile where reliability degrades as task complexity increases. We address the challenge of generating high-quality natural language executable plans for complex tasks by

Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents

arXiv:2609.26760v2 Announce Type: replace Abstract: Large language model (LLM) agents often handle streams of related tasks, yet standard harnesses repeatedly ask the model to reconstruct the same control decisions inside each task's context. We study whether task feedback can in

Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction

arXiv:2609.25176v2 Announce Type: replace-cross Abstract: Real-time voice assistants must reason over evolving requests, execute actions, and follow conversational rules. Qwen-Audio-3.1-Realtime brings these requirements together through Think, Act, and Speak and Coordinate. Thin

Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents

arXiv:2609.13422v2 Announce Type: replace Abstract: LLM judges are increasingly used to evaluate and improve AI-generated outputs, yet their reliability for complex professional work remains unclear. We study this problem through Vibe Patenting, an end-to-end patent-drafting test

Where Does Exactly-Once Live? Model, Harness, and Tool-Contract Effects on Duplicate Side Effects in LLM Agents

arXiv:2609.29095v1 Announce Type: cross Abstract: When a tool-using agent's write times out or returns a server error, the action may already have taken effect. Retrying blindly duplicates it -- a second charge, a second announcement, a second deployment -- while giving up skips