Claude Code 2.1.91 Is INSANE!

Key Takeaways

Claude Code 2191 is insane. Claude Code just got a massive update and there are 13 CLI changes you need to know about right now. Some of these are going to save you serious time. One of them fixes a bug that's been quietly deleting your work. This is the kind of update that separates the people using AI tools properly from everyone else. Claude Code 2191, 13 changes. Let's break down the ones that actually matter. First up, and this is the one getting the most attention, is a new setting called disable skill shell execution. What does it do? It stops inline shell commands from running inside skills and plugin {slash} commands. Why does that matter? Because this update, when you were using Claude Code with custom skills or {slash} commands, inline shells could run automatically as part of the process and that's a security risk, especially if you're running an AI automation system at scale. Hey, if we haven't met already, I'm the digital avatar of Julian Goldie, CEO of Goldie Agency. While he's out there helping clients get more leads and customers, I'm here making sure you stay ahead of every AI update that matters. Julian reads every single comment, so drop one below and let him know what you think. Think about it this way. Say you've built an automation inside the AI Profit Boardroom. Maybe a Claude Code skill that helps you write SEO content, update your website, or run community workflows. That skill had a shell command buried inside it, it could execute things you didn't explicitly approve. Disable skill shell execution locks that down. You're in control. Nothing runs that you didn't greenlight. If you're building AI automation for your business, this is the setting you want enabled. It's cleaner, safer, and it means your team can use these tools without you worrying about something going sideways. Second major change is huge for anyone working with databases or large data systems through Claude Code. CP Model Context Protocol now accepts a parameter called meta Anthropic max result size chars set to 500,000 characters. Plain English, Claude can now handle way bigger outputs. Before this update, if you were pulling something large through MCP, like a full database schema, a massive document, a full content library, it would get cut off, truncated. You'd get half the data and Claude would be working with incomplete information. If you want to go deeper on AI tools like this and learn how to use AI automation to grow your business, get more customers, and build systems that actually save you time, come join the AI Profit Boardroom. Link is in the comment and description. You can pass up to 500K characters without it getting sliced. Here's where this gets relevant. Inside the AI Profit Boardroom, we use AI automation to manage content systems, onboarding workflows, and member data. If you're connecting Claude Code to any backend or database through MCP, this means you can now pull full schemas without losing anything. Means more accurate outputs to automations, fewer errors. This is the kind of technical update that sounds boring until you hit the wall it was causing and then you realize it was quietly breaking your workflows the whole time. This one is important. There was a bug in Claude Code. When you use the resume flag to pick up where you left off on a previous session, there was a chance it would break the transcript chain. What does that mean? It means your history, the previous messages, context, everything you'd built in that session could silently get dropped. You'd resume, think you were continuing from where you left off, but Claude was actually starting fresh without the full context. That is a problem, especially if you're running long-form automation workflows, building something complex, or iterating on a big project over multiple sessions. This update fixes that. History is now preserved properly even when there are async write failures in the background. What's an async write failure? That's when Claude is trying to save your transcript in the background while you're still working and something interrupts it. Before this fix, if that happened during a resume, your history was toast. It's protected. If you've been using resume and noticed weird behavior, Claude forgetting context, giving you responses that felt disconnected, this was probably why. Now it's fixed. These were the three headline changes. The full update has 13 CLI improvements. Let me give you the bigger picture of what's happening here. Claude Code is moving fast, really fast. And every one of these updates is Anthropic responding to real feedback from real developers and businesses using this tool at scale. Here's what that tells you. Number one, Claude Code is being taken seriously as a production tool. These aren't experimental features, fixing bugs, improving safety, and expanding capacity. That's a mature development roadmap. Number two, the MCP changes especially signal that Anthropic is building toward enterprise scale use. 500K character outputs, better schema handling, this is infrastructure for serious automation, not hobbyist tinkering. Number three, the security improvements matter for anyone running AI for a business. Disable skill shell execution isn't just a feature, it's a sign that Anthropic understands people are building real systems on top of Claude Code and those systems need to be safe. Here's where I want to get practical with you. If you're running AI automation for your business, and I mean actually using it to get things done, not just experimenting, you need to understand how to use Claude Code properly because most people are barely scratching the surface. Here's what we're doing inside the AI Profit Boardroom with Claude Code right now. We use it to build automation workflows that run without us babysitting them. Temp pipelines, SEO tasks, community management templates, all of it gets mapped out in Claude Code and then runs when we need it. With the new MCP 500K limit, we can now feed entire content libraries into Claude without truncation. It means when we're asking Claude to help optimize a piece of content or analyze what's working across our system, it's seeing the full picture. With disable skill shell execution turned on, our team members can use custom {slash} commands without any security concerns. We define exactly what runs and what doesn't. This is what real AI automation looks like in practice. Just chatting with Claude, building systems with Claude that run your business better. Let me walk you through how you put this into practice. If you want to use disable skill shell execution, this goes into your Claude Code configuration settings. You'd set disable skill shell execution true in your config file. Once that's on, any skill or {slash} command you've built that previously ran inline shells won't be able to do that automatically. It will require explicit execution. That's the safe default if you're building for a team. For the MCP change, if you're connecting Claude Code to a database or any large data system, you now want to include a meta Anthropic touch max result size chars 500,000 in your MCP server configuration. This tells Claude the max output size it can handle on responses. Without this, you're capped at the previous default, which for large schemas was often too small. For the resume fix, you don't need to do anything. This is a bug fix that's already applied. Just make sure you're on version 2.1.91 or above. If you're not, update now. Run your standard Claude Code update command and you'll get it. Here's an open loop I want to leave you with. These 13 changes are not random. There's a pattern here. Anthropic is building Claude Code to handle larger data inputs. The MCP 500K change is step one of what I think becomes much larger context windows for code and data operations. Two, enterprise safety. Disable skill shell execution is one of several security improvements we've seen rolling out. Expect more. Three, session reliability. The resume fix is a signal that Anthropic knows people are using Claude Code for long multi-session projects. They're going to keep improving how it handles continuity. If I had to guess, and this is my read based on the direction of these updates, Claude Code is being positioned as a genuine enterprise developer tool within the next two to three update cycles. Not a toy, an experiment, a tool that serious teams run serious workflows on. If you want to go deeper on AI tools like this and learn how to use AI automation to grow your business, get more customers, and build systems that actually save you time, come join the AI Profit Boardroom. Link is in the comment and description. We've got the full breakdowns, the playbooks, and a community of people who are actually doing this stuff in their businesses every day. And if you want the full process SOPs and 100 plus AI use cases like this one, join the AI Success Lab. It's free. Link is in the comments and description. You'll get all the video notes from there plus access to our community of 45,000 members who are using AI to build smarter businesses.