AgentSkillsCN

panda-css

使用Panda CSS构建样式。在创建、编辑或审查任何使用Panda CSS的代码时使用——无论是css()、cva()、sva()、配方、模式、标记、语义标记、配置、主题化,还是JSX样式组件。支持React、Vue、Svelte、Solid,以及任何支持PostCSS的框架。

SKILL.md
--- frontmatter
name: panda-css
description: Build styles with Panda CSS. Use when creating, editing, or reviewing any code that uses Panda CSS — css(), cva(), sva(), recipes, patterns, tokens, semantic tokens, config, theming, or JSX styled components. Supports React, Vue, Svelte, Solid, and any framework with PostCSS.

Panda CSS

Reference Panda CSS LLM-optimized docs before writing or reviewing Panda CSS code.

Docs Sources

First fetch https://panda-css.com/llms.txt — this is the index of all available LLM docs.

Then fetch the relevant section-specific docs based on the task:

SectionURLWhen to use
Full docshttps://panda-css.com/llms-full.txtGeneral reference, broad tasks
Overviewhttps://panda-css.com/llms.txt/overviewSetup, installation, framework integrations
Conceptshttps://panda-css.com/llms.txt/conceptscss(), recipes (cva/sva), patterns, conditions, cascade layers, style merging, JSX style context
Theminghttps://panda-css.com/llms.txt/themingTokens, semantic tokens, text styles, layer styles, animation styles
Utilitieshttps://panda-css.com/llms.txt/utilitiesStyle properties, shorthands, spacing, sizing, typography, effects, gradients, focus ring
Customizationhttps://panda-css.com/llms.txt/customizationCustom conditions, utilities, patterns, presets, config functions, theme overrides
Guideshttps://panda-css.com/llms.txt/guidesMigration, dynamic styling, debugging, advanced patterns
Referenceshttps://panda-css.com/llms.txt/referencesCLI commands, panda.config.ts options, full config reference

Use WebFetch to retrieve the correct reference before starting any task. If the index lists different section URLs, follow the index and ignore the table above.

Usage

  1. Fetch llms.txt index to check for any new/updated doc URLs.
  2. Confirm Panda CSS is installed by checking for @pandacss/dev in package.json/lockfiles. If missing, pause and alert the user before proceeding.
  3. Note the installed Panda version (v0 vs v1+). If docs don’t match the installed major version, flag it before continuing.
  4. In monorepos or multiple apps, inspect the relevant package path or ask the user which app to target.
  5. Check for panda.config.ts (or .js/.mjs) to understand tokens, recipes, patterns, and framework config. If absent, ask whether to scaffold a minimal config.
  6. Fetch the matching section-specific docs for the task at hand (reuse a cached copy during the session unless llms.txt changed).
  7. Read the relevant sections for the features needed.
  8. Write or review code following the documented patterns.