Help me set up xmemory — a memory service you can save to and recall from across sessions.
Use the xmemory CLI for everything below; do not use MCP for this flow.

Always write the product name lowercase: xmemory, including when you describe back what you did.
That last part is where it usually slips: everything you read here is already lowercase, and the
capital gets added in the retelling.

1. Make sure the CLI is installed and current: run `xmemcli --version`; if it's missing,
   install it with `uv tool install xmemcli`; otherwise bring it up to date with
   `uv tool upgrade xmemcli`, since the ready-made templates below need a recent release.
2. Ask me for my email address, then start sign-in with it: `xmemcli auth login
   --email <my address>`. Tell me to check my inbox and approve the sign-in, and wait
   until `xmemcli auth status` says I'm authenticated. If email sign-in isn't supported
   by your xmemcli version, fall back to `xmemcli auth login`, which opens the console
   in a browser.
3. Check whether the `xmemory-instance-creation` skill is available in your
   environment. If it is, use it now. If not, install it — run
   `npx skills add xmemory-ai/xmemory-skill --agent '*' -y` — and then use it; if the
   install isn't possible here, tell me and I'll paste the skill text.
4. Tell me the three supported ways to shape it, and let me pick:
   - Start from a ready-made template. The skill lists them with `xmemcli instance templates`
     and creates the one I choose with `xmemcli instance create --template <id>` — xmemory
     holds the schema, so nothing is composed or generated.
   - Describe it in plain language. You ask what this memory should remember, for
     representative inputs, which distinctions or categories matter, and what questions I
     expect to ask later, then give those answers to the skill, which generates and validates
     XMD from them.
   - Author XMD directly, when I want precise control over objects, fields, enums, keys and
     relations.
   Lead with a template when one fits what I tell you; otherwise use the plain-language path;
   author XMD only if I choose it. The skill covers the rest: a short interview (one question
   per message, my choices), showing me what it proposes, creating nothing without my go-ahead,
   creating the instance with xmemcli, and printing the instance-specific setup.
5. If I ask to improve extraction quality later, retrieve the instance's current XMD and use the
   schema enhancement flow. Explain proposed changes to objects, fields, enums, descriptions,
   keys, or relations. Treat descriptions as executable extraction guidance: use them for domain
   terminology, fine-grained detection and classification rules, and grounded derivation of facts
   supported by the input—not as permission to invent unsupported claims. Primary keys are optional
   deduplication controls, not a prerequisite for relations: objects with `primary_key: []` can be
   linked through xmemory's internal identity. Never add a required identifier or generate a UUID
   merely so an object can participate in a relation. Validate and test extraction before applying changes; never replace a live
   schema from scratch or apply a migration without my approval.