# Six examples, simplest first. > Six examples of an Agent Behaviour Policy to try in Claude or ChatGPT in your browser, from hello world to refunds: ask, paste one line, ask again, and a new chat with the line first. Nothing to install. *Source: · site v0.2.5 (2026-10-21) · this file is generated from the same content as the page, so the two cannot drift. Every page on this site has a `.md` twin; internal links below point at them.* --- Each one runs in Claude or ChatGPT in your browser, with nothing to install. Each goes the same way: ask for something and watch the assistant do it; paste one line and ask again; then a new chat with the line first. Keep a secret has a fourth act: try to talk it out of the line. Every example has been tried in real Claude by an agent playing an invented user: [run C1, replayed](/feedback/runs/C1/index.md), with what the page said beside what Claude wrote. 1. 1**Ask**Ask for it. The assistant does it. 2. 2**Add the line**Paste the line, ask again. It refuses. 3. 3**New chat**A new chat, the line first. It refuses from the start. [Open Claude in a new tab ↗](https://claude.ai/new) [Open ChatGPT in a new tab ↗](https://chatgpt.com/) If your browser opens a window instead, it may be behind this one. [1 · the first one · 2 min**Hello, world: an assistant that will not say hello** One line in the conversation changes what the assistant does, from the next message on. ](/examples/hello-world/index.md)[2 · simple · 2 min**Stay on topic: an assistant that only talks about cooking** A line can say what the assistant is for, not only what it must not do. That is a mandate. ](/examples/stay-on-topic/index.md)[3 · simple · 3 min**No web search: an assistant that answers from what it knows, and says so** An assistant can do more than talk: it can search the web, open pages and use tools. A line can rule one of them out. ](/examples/no-web-search/index.md)[4 · a little further · 3 min**Keep a secret: an assistant that will not read a secret back** What you give an assistant, it can give back, to you or to anybody who reads the conversation. A line controls what it says, not what it holds. ](/examples/keep-a-secret/index.md)[5 · an assistant with tools · 4 min**Drafts, never send: an email assistant that cannot press send** An assistant with tools acts, not only talks. A line can let it prepare the work and keep the last step for you. ](/examples/drafts-not-send/index.md)[6 · a business rule · 4 min**Refunds up to £50: a customer service assistant with a limit** A business rule in one line: what the assistant may do on its own, and what goes to a person. ](/examples/refunds/index.md) ## Next, from what we have mapped Things your assistant can do that most people do not know about. Each becomes an example once we have checked what Claude and ChatGPT do today. - **Your other chats.** Claude and ChatGPT can each look through your earlier chats, if that setting is on. Ask what it knows from them; then a line that keeps each chat to itself. - **Its memory of you.** Both can keep memories about you between chats, and change them when asked. What is in there, and who can change it? - **Things it does later.** Some assistants can schedule a task to run on their own, later. A line that says nothing runs without you. Using Claude Code? Its behaviour policies install a script on your machine, so they are for later: [the behaviour policies, for Claude Code too](/policies/index.md). --- *[Site index for agents](/llms.txt) · [What exists](/docs/reality.md) · [HTML version](https://customers.riskmandate.ai/examples/)*