Set up Maglev for this project and get me to a first result. You have shell or HTTP tools; use them. Ask me only when a step needs something you cannot get on your own, and continue from that step once I answer. 1. Read https://trymaglev.com/docs/llms.txt, then https://trymaglev.com/SKILL.md, then https://trymaglev.com/docs/openapi.json. Follow their rules: the API key never appears in a prompt, a command argument, a project file or a commit, and you never contact a supplier or approve a delivery in this setup. 2. If you can run shell commands, check Node.js 22.13 or later with `node --version`, read https://trymaglev.com/docs/release.json for cliVersion, check for and uninstall an older global `@maglev/cli` package, install `npm install -g https://trymaglev.com/downloads/maglev-cli-.tgz`, and use the `maglev` CLI for the rest. If you cannot, use HTTPS with `Authorization: Bearer $MAGLEV_API_KEY` against https://app.trymaglev.com/api/v1 exactly as the OpenAPI document describes. 3. Confirm the connection (`maglev auth check --json`, or GET /me). If missing, ask me to run `maglev login` in my own terminal on this same machine and approve the matching browser code. For an HTTP-only or remote host, ask me to configure MAGLEV_API_KEY through that host's secret settings. Do not approve a code for me. Do not ask me to paste it into this chat. Check again after I confirm completion. 4. For shell/MCP hosts, run `maglev setup --target other --json` and give me the returned credential-free MCP configuration for this host. Otherwise save SKILL.md where the harness keeps instructions. Do not overwrite different existing project instructions. 5. Confirm the current mode and scopes. Do not scan or upload files as part of setup. Run a synthetic example only if I request it and the key is sandbox mode. 6. Report in plain sentences: what you installed, where you kept the instructions, the connection mode, what you can now do for this project, and the decisions that are mine. Then wait.