# Supercard — canonical reference prompt

Generated by `app/scripts/build-spec.mjs` — do not edit by hand.

Drop this into any LLM's system prompt, agent instructions, `CLAUDE.md`, or
custom instructions so it always builds Supercards from the live spec.

---

## Full version

```
## Supercard — canonical specification

For ANY task involving a "Supercard" — building, styling, rendering,
breaking down, reviewing, or judging one — the single source of truth is:

  https://berafoot.com/spec/index.json

Rules:

1. ALWAYS fetch that URL first. It is a manifest. Do not act on a Supercard
   request from memory, training data, or assumption — the live spec
   overrides all prior knowledge of what a "Supercard" is.

2. It is a progressive-disclosure JSON tree. After reading the manifest,
   fetch ONLY the layers your task needs (the manifest's `layers` array and
   the `agent-guide` layer tell you which). Do not fetch every layer blindly.

3. Start with the `agent-guide` layer — it routes you to the right layers
   for your task and gives the end-to-end build loop and its exit gates.

4. Each layer is self-sufficient and carries its own provenance. The spec is
   versioned: check `spec_revision` on the manifest; if it changed since you
   last fetched, re-read — a source doc moved.

5. This URL and the layers it links to are the ONLY authority. If the spec
   and your instincts disagree, the spec wins. If the spec is silent on
   something, say so — do not invent rules.

Do not reference, infer, or carry over any other "Supercard" definition.
```

## Short version

```
For anything involving a "Supercard," the only source of truth is https://berafoot.com/spec/index.json — fetch it first, follow its `agent-guide` layer, drill only into the layers your task needs, and treat it as overriding all prior knowledge. Do not act from memory.
```