---
name: <skill-name>
# The description is the ONLY text always in context — it decides when the skill fires.
# Write it as a clear trigger: what the skill does + when to use it. Tune for precision.
description: <what this skill does and when to use it — e.g. "Reconstruct a prototype screen 1:1 in the project's UI stack. Use when building or wiring a screen from the rendered prototype.">
# Optional (Claude Code): restrict tools, or fork into an agent type.
# allowed-tools: Read, Grep, Glob, Edit, Write, Bash
---

> **Specialization:** generic by design — bind every `<…>` and "per `CLAUDE.md`" reference from the project `CLAUDE.md` (stack, design-system prefix, naming grammar, domain compliance). See `SPECIALIZE.md`. If a needed value is missing from `CLAUDE.md`, STOP-and-ask.

## What this skill is for
<!-- 1–2 sentences. The capability this teaches and the situations it applies to. Mirror the description. -->

## When to use it
<!-- The triggers. Be specific so the model invokes it at the right moment and not others. -->

## How to do it
<!-- The procedure, as concise steps. Keep this body to about a screenful. -->
1. <step>
2. <step>
3. <step>

## Rules / guardrails
<!-- The non-negotiables this expertise enforces (the "always / never" that make output correct). -->
- <rule>

## Reference (progressive disclosure)
<!-- Push long detail into files the body points to, so they load only when needed. -->
- `reference/<topic>.md` — <what it contains; loaded on demand>

## Done when
<!-- The check that the work this skill guided is actually correct. -->
- <verifiable condition>