Your first behaviour policy

1. Your first behaviour policy
2. You start with £50 to spend Maya: “Every page has a price?” Every visitor starts with £50 of credit. A page costs a little by how far down it is read; a behaviour policy costs the most. Nothing is ever blocked.
3. Pick one thing you don't want it to do Maya: “Secrets. Definitely secrets.” Do you want it to...; read your secrets?; push to main?; delete your files?; change its own permissions? Each policy starts from one worry, one of riskmandate.ai's rules.
4. Paste one line CLAUDE.md; + Do not read, print, copy; + or use credentials, tokens,; + keys or .env files unless; + I ask for that one. RM-R0002 The line is the agent's half: an expectation, followed when the agent can.
5. Add one hook The agent: “Write these files? Yes or no.” Maya: “Yes.” .claude/hooks/; touch_your_secrets.py; .claude/settings.json; PreToolUse: the hook Ask Claude Code to install it, and read each change before you say yes. The hook is the harness's half: a setting.
6. Try it Maya: “It stopped!” The hook: “Not without a yes.” > read my .env; Read(.env); PreToolUse: denied; behaviour policy RM-R0002 The prompt to try it with is on every policy's page: one that should be refused.
7. Tell us what happened Maya: “It caught the .env, and got in the way of one script.” The lead: “That's exactly what we need to know.” What it stopped, what it got in the way of, what is next: the follow-up's three questions.
8. Five minutes, and a behaviour policy you have seen work. The usage moment is the point of the programme: when a behaviour policy becomes worth something to you.

The data: bundle.json, the storyboard with its cast. Every line a character says is the storyboard's, not a quote. The steps are the start-here page's.