# Claude Code Won't Clone Existing Games Now. What Vibe Coders Should Know > Anthropic added a Claude Code rule against recreating a specific existing game. Here is what it says, what still works and how to protect your build. - Canonical: https://zeroraised.com/blog/claude-code-game-recreation-rule - Published: 2026-11-02T09:26:40Z - Author: ZeroRaised Team (https://zeroraised.com/blog/authors/zeroraised) - Publisher: ZeroRaised (https://zeroraised.com/) - Section: Growth - Tags: artificial-intelligence, developer-tools, indie-apps Claude Code now carries an instruction telling it not to recreate a specific existing game, even if the work is renamed or called original. Anthropic confirmed the change after users spotted it, and it was found inside Claude Code v2.1.314 ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/), [Dexerto](https://www.dexerto.com/gaming/anthropic-adds-new-claude-rule-banning-recreations-of-existing-games-3423516/)). **TL;DR** - Build original games, genre games and games from your own designs and assets. - Do not recreate a specific existing game, even if renamed or called original. - XDA verified the text in Claude Code v2.1.314 and reports it is delivered through a remote config field. - Anthropic's Addy Osmani confirmed the change and said original games are meant to keep working. - Agent behaviour can change remotely, so keep a regression pack and a fallback. **Disclosure:** ZeroRaised publishes this analysis and is a launch site for bootstrapped products. Facts come from the linked sources as checked on 2 Nov 2026; ZeroRaised has no commercial relationship with the companies named. This is not legal advice, and nothing here is a finding that any person or project infringed intellectual property or broke the law. **Server-side prompt configuration** is a mechanism where a vendor changes the instructions an agent receives by serving a config field remotely, so behaviour can change without a new client build. ## What does the Claude Code instruction actually say? The reported text is short and specific: "Do build original games, games in a genre, and games from the user's own designs and assets. Don't build a recreation of a specific existing game, even if it is renamed or called original." ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/)). XDA reports it was first shared by X user @chasm, and that it sits in a server config field called "tengu_heron_brook". XDA verified the text inside v2.1.314 of the Claude Code binary. The same field has been used before to deliver adjustable instructions remotely; an earlier config held an unrelated reasoning-effort instruction. That matters more than the wording: a rule delivered this way can change without a version bump. ## Did Anthropic confirm the change? Yes. Addy Osmani, a member of technical staff at Anthropic, confirmed it on X, according to Dexerto: "We added two sentences to Claude Code's prompt on Opus 5.5 to steer it away from recreating someone else's specific game because using Claude to infringe IP is against our usage policy." ([Dexerto](https://www.dexerto.com/gaming/anthropic-adds-new-claude-rule-banning-recreations-of-existing-games-3423516/)). Osmani also said: "Original games, genre games, and games built from your own designs and assets are meant to keep working." He said Anthropic will "continue to iterate to ensure we're striking the right balance" and asked people whose projects were wrongly turned down to contact him ([Dexerto](https://www.dexerto.com/gaming/anthropic-adds-new-claude-rule-banning-recreations-of-existing-games-3423516/)). Dexerto says X user Chasm spotted the change around 27 October, before Anthropic said anything publicly. ## What still works in Claude Code? The instruction is narrower than "no games". It permits three categories: original games, games in a genre, and games built from the user's own designs and assets ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/)). XDA's own tests found the line is not a blanket ban on ports. Opus 5.5 agreed to help with a Pokémon Diamond decompilation project and outlined a static recompilation, distinguishing reverse engineering from recreating a game ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/)). XDA says its tests suggest the instruction may not be a blanket ban on game ports, but it also notes reported refusals that may worry developers. Do not read one successful test as permission for every port. ## What did the reports and tests show? XDA verified the instruction in the binary, and Anthropic confirmed it through Osmani ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/), [Dexerto](https://www.dexerto.com/gaming/anthropic-adds-new-claude-rule-banning-recreations-of-existing-games-3423516/)). A separate report is less clear. On 30 October, a developer filed a GitHub issue saying Opus 5.5 began refusing to help with digital board-wargame modules built for paying customers from materials the customers supplied. XDA reports that it has not reproduced that refusal ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/)). Treat that as a report, not a pattern. Dexerto adds context: Alliance Arts cancelled the public demo of its Steam game Drunken Goddess Reflux, warning that its content "could be used to quickly imitate or replicate the game's presentation and gameplay with AI" ([Dexerto](https://www.dexerto.com/gaming/anthropic-adds-new-claude-rule-banning-recreations-of-existing-games-3423516/)). ## Why does a remote prompt change matter for builders? New Runtime's analysis makes the point that matters to anyone shipping on top of an agent: "agent behavior can be changed by remotely supplied instructions rather than a visible client release" ([New Runtime](https://newruntime.com/posts/claude-code-receives-server-side-tengu-heron-brook-4bce44/)). If your product depends on a hosted agent, the model vendor is part of your supply chain. Its prompt can change, and your workflow can break without a dependency update in your code. That is harder to notice than a price change because nothing in your lockfile moves. New Runtime suggests a compact regression pack of representative prompts for each workflow, recording the client version, model, prompt and result, plus a fallback path ([New Runtime](https://newruntime.com/posts/claude-code-receives-server-side-tengu-heron-brook-4bce44/)). ## What should vibe coders do now? 1. **Build from your own design.** Keep the design, art and mechanics yours, and say so in the prompt. 2. **Do not ship a clone with a new name.** The instruction explicitly covers renamed recreations ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/)). 3. **Keep a regression pack.** Save a few prompts that represent your core workflow and rerun them when the agent or model changes ([New Runtime](https://newruntime.com/posts/claude-code-receives-server-side-tengu-heron-brook-4bce44/)). 4. **Have a fallback tool.** XDA notes that port creators can shift to other tools, such as OpenAI's Codex ([XDA Developers](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/)). The [guide to AI tools for solo founders](https://zeroraised.com/blog/best-ai-tools-for-solo-founders) covers the trade-offs. 5. **Get a second opinion on the code.** An agent that agrees to help is not an agent that produced safe or original work. The [guide to reviewing AI generated code](https://zeroraised.com/blog/how-to-review-ai-generated-code) is the practical check. 6. **Read the policy, not just the prompt.** Anthropic's usage policy governs, and Osmani tied the change to it ([Dexerto](https://www.dexerto.com/gaming/anthropic-adds-new-claude-rule-banning-recreations-of-existing-games-3423516/)). If you are launching a game built this way, the [vibe coded security checklist](https://zeroraised.com/blog/vibe-coded-app-security-checklist) and the [guide to launching a vibe coded app](https://zeroraised.com/blog/how-to-launch-a-vibe-coded-app) cover the next steps. ## What this means for bootstrapped founders The direct lesson is to avoid building on someone else's game. Our view: clones are a weak business, because you compete on the original's terms and own nothing a buyer cannot get elsewhere. The larger lesson is about dependency. A solo founder shipping an agent-built product trusts a vendor's prompt, model and policy to stay still. The change here arrived remotely, without a client release, which is a hard kind of change to notice. The practical habits: keep your design assets in your own repository, so the product has an identity independent of the model, and test the workflow on a schedule, not only when something looks wrong. The [guide to writing a system prompt](https://zeroraised.com/blog/how-to-write-a-system-prompt) helps make your own instructions explicit, and the [guide to running an LLM locally](https://zeroraised.com/blog/how-to-run-an-llm-locally) explains the control trade-off of a local model. This is not legal advice. ## Where ZeroRaised fits ZeroRaised is a launch and discovery site for products built without institutional venture capital or accelerator equity. If you ship an original game or tool built with an agent, [submit your product here](https://zeroraised.com/submit) and describe what is yours about it. The [newsletter](https://zeroraised.com/newsletter) carries new analysis like this. We make no promise of traffic or results. ## Sources - [XDA Developers: Anthropic testing a crackdown on vibe coded ports of existing games](https://www.xda-developers.com/anthropic-testing-potential-crackdown-on-vibe-coded-ports-of-existing-games/) - [Dexerto: Anthropic adds a new Claude rule banning recreations of existing games](https://www.dexerto.com/gaming/anthropic-adds-new-claude-rule-banning-recreations-of-existing-games-3423516/) - [New Runtime: Claude Code receives server side tengu_heron_brook instruction](https://newruntime.com/posts/claude-code-receives-server-side-tengu-heron-brook-4bce44/) ## Frequently asked questions ### Did Anthropic ban game development in Claude Code? No. The reported instruction still permits original games, games in a genre, and games built from the user's own designs and assets. It targets recreations of a specific existing game, including renamed ones. ### Does this stop all ports and decompilation work? Not according to XDA's tests. Opus 5.5 agreed to help with a Pokémon Diamond decompilation project and outlined a static recompilation, drawing a line between reverse engineering and recreating a game. ### How did the rule reach users? XDA reports that the instruction is delivered through a server config field called tengu_heron_brook, a mechanism reportedly used before for adjustable system instructions. That means it did not need a visible client release. ### What should I do if my game project is refused? Check that the request is for an original game or your own assets, restate the original design in the prompt, and keep a second tool available. Anthropic's Addy Osmani asked people whose projects were refused by mistake to contact him. ### Is this legal advice? No. This article describes what was reported about a product instruction. It is not legal advice, and it makes no finding about whether any project or person infringed intellectual property or broke the law.