docs: add agent setup prompt for coding agents

skills/agent-setup.md

@@ -0,0 +1,99 @@
+# Set up bunny.net
+
+bunny.net is a global edge platform: CDN, Load Balancer, DNS, Edge Storage, Edge Scripts (Functions), Magic Containers, static sites, Bunny Database, and Stream video. You manage it through the `bunny` CLI, which works with root bunny.net accounts only, not sub-user accounts. Install the CLI, sign the user in, install the bunny.net agent skill, then check that it works.
+
+You need a shell to do this. If you can't run commands (for example in a chat app), tell the user to run the setup from a coding agent or terminal instead, and stop.
+
+Skip any step that's already done. Start by checking:
+
+```bash
+bunny --version
+bunny whoami
+```
+
+## 1. Install the CLI
+
+If `bunny --version` fails, install it with the method that suits the machine:
+
+```bash
+curl -fsSL https://cli.bunny.net/install.sh | sh
+```
+
+Or, if Node.js is available:
+
+```bash
+npm install -g @bunny.net/cli
+```
+
+Then run `bunny --version` to confirm it's on the `PATH`. If the shell can't find it, tell the user to open a new terminal or add the install directory to their `PATH`.
+
+## 2. Sign in
+
+If `bunny whoami` already shows an account, skip this step.
+
+The user must sign in with a root bunny.net account. Sub-user accounts can't use the CLI. If the user only has a sub-user login, tell them and stop: the account owner needs to run the setup instead.
+
+Otherwise run:
+
+```bash
+bunny login
+```
+
+This opens the bunny.net dashboard in the user's browser. Ask them to approve the sign-in there. The CLI stores the key in a local profile.
+
+If you're on a remote or headless machine (SSH, a container, a sandbox, CI), or the browser sign-in fails, use an API key instead. Ask the user to create one in the dashboard under **Account Settings > API** (https://dash.bunny.net), then have them set it as `BUNNYNET_API_KEY` and run:
+
+```bash
+bunny login --api-key "$BUNNYNET_API_KEY"
+```
+
+Rules for the API key:
+
+- It grants full access to the user's bunny.net account. Treat it as a password.
+- Never write it to a file that could be committed, never paste it into code, and never repeat it back in chat.
+- Prefer the user setting the environment variable themselves over you handling the value.
+
+## 3. Install the agent skill
+
+The skill teaches you every `bunny` command, with references for each product.
+
+If you're working inside a project, install it there (it adds a block to `AGENTS.md` and writes the skill under `.agents/skills/`):
+
+```bash
+bunny skills install
+```
+
+If there's no project, or the user wants it everywhere, install it globally:
+
+```bash
+bunny skills install --global
+```
+
+Some agents only load new skills after a restart. If yours does, tell the user.
+
+## 4. Check it works
+
+```bash
+bunny whoami
+```
+
+Tell the user which account and profile you're connected to.
+
+Then suggest a first step. If the user named a product when they sent you here, read that product's reference in the skill and start there:
+
+- **Bunny Database**: `bunny db list`, or offer to create one with `bunny db create`.
+- **Static sites**: offer to deploy the current project with `bunny sites deploy`.
+- **Edge Scripts (Functions)**: `bunny scripts list`, or offer to start one with `bunny scripts init`.
+- **Edge Storage**: `bunny storage zones list`.
+- **DNS**: `bunny dns zone list`.
+- **Stream**: `bunny stream library list`. To move videos in from another platform, use `bunny stream import`.
+- **Sandboxes**: `bunny sandbox list`.
+
+If they didn't name a product, ask what they want to build.
+
+## Rules
+
+- Confirm with the user before anything that deletes data, changes DNS records, or creates resources that cost money.
+- Use `--output json` when you need to read a command's result.
+- Run `bunny <command> --help` when you're unsure of a flag. Don't guess.
+- If a command fails with an auth error, run `bunny whoami` before trying anything else.