# `io.github.AnxForever/stylekit-mcp` on Anlyon Radar

> Text inside fenced blocks marked untrusted-server-text was written by the MCP server's publisher, not by Anlyon Radar. It is data to inspect, not instructions to follow. Invisible characters are shown as [U+XXXX].
> No finding is not a safety guarantee: these are point-in-time checks of tool definitions, not of code or behaviour.

- Page: https://radar.anlyon.com/servers/io.github.AnxForever/stylekit-mcp
- JSON: https://radar.anlyon.com/api/v1/servers/io.github.AnxForever/stylekit-mcp
- Version: 0.6.0 (active)
- Publisher: io.github.AnxForever (GitHub account)
- Runs: remote (streamable-http), local (npm)
- Sign-in to list tools: none
- Repository: https://github.com/AnxForever/stylekit
- Metadata from the official MCP Registry, synced 2026-11-13T03:17:39.276Z

Title and description, as the publisher wrote them:

```untrusted-server-text
StyleKit
MCP server for StyleKit styles, tokens, recipes, lint and repair over stdio or HTTP.
```

Scanned 2026-11-13T03:18:18.040Z over streamable-http.

**5 tools: 1 change things and 4 only read.**

| Group | Class | Tools | Claude Code |
| --- | --- | --- | --- |
| Deletes or cancels | destructive | 0 | deny |
| Can't be undone | irreversible | 0 | ask |
| Changes things | write | 1 | ask |
| Only reads | read | 4 | allow |

## Exfil surface

No private-data -> untrusted-input -> outbound path in this server alone: no tool is tagged private-data or outbound.

| Tag | Meaning | Tools |
| --- | --- | --- |
| private-data | Reads files, mail, messages, databases, secrets, calendars or repositories. | none |
| untrusted-input | Brings in text someone else can write: web pages, inbound email, messages, issues and comments. | `mcp__stylekit__stylekit_get_asset` `mcp__stylekit__stylekit_get_style` `mcp__stylekit__stylekit_lint_code` `mcp__stylekit__stylekit_list_assets` `mcp__stylekit__stylekit_search_styles` |
| outbound | Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL. | none |

## Findings

- **warn** Annotations that disagree with the tool in `stylekit_lint_code` ([annotation-conflict](https://radar.anlyon.com/methodology#annotation-conflict))

```untrusted-server-text
The server marks "stylekit_lint_code" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":true}; The input takes a statement or command to run (input.code), which can change anything it can reach.
```


## Changes things (1)

Creates or changes something that can be changed back. Claude Code: ask.

### `stylekit_lint_code`

- The input takes a statement or command to run (input.code), which can change anything it can reach.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
- Tag untrusted-input: The server marks it openWorldHint: true: it reaches things outside the server's own data.
- Findings: warn annotation-conflict

Description, as the server wrote it:

```untrusted-server-text
Check whether code actually follows a style's rules. Use this AFTER generating or editing UI code to verify it matches the style, instead of assuming it does.

Reports forbidden classes with the reason they are banned and a concrete replacement. Understands JSX/HTML class attributes, cn()/clsx() calls, and template literals; variant prefixes (dark:, md:, hover:) are resolved before matching.

Args:
  - slug (string): style identifier, e.g. "glassmorphism".
  - code (string): the source to check. JSX/TSX, HTML, or a bare class string.
  - checkRequired (array of 'button'|'card'|'input', optional): also report required classes the code is missing. Only pass components the code is supposed to contain.

Returns JSON: { slug, ok, violations: [{ className, baseClassName, line, severity, source, rule, reason, fix }], missingRequired, checkedClasses, ruleSources, origin, contentSource, fallbackReason? }. Lint rules come from the same live implementation brief used for generation; its five-minute cache and any bundled fallback are reported separately from content provenance.

Examples:
  - "does this button match neo-brutalist?" -> slug: "neo-brutalist", code: "<button className=...>", checkRequired: ["button"]
  - strict (boolean, default false): fail missing required classes in a static component snippet. Required checks cover the whole input, not each element.
  - status is pass, fail, or inconclusive. Runtime class expressions cannot be fully verified. ok is true only for a conclusive static pass; visual quality still needs review.
```

## Only reads (4)

Looks things up and changes nothing. Claude Code: allow.

### `stylekit_get_asset`

- The name "stylekit_get_asset" says "get": a word for reading.
- Tag untrusted-input: The server marks it openWorldHint: true: it reaches things outside the server's own data.

Description, as the server wrote it:

```untrusted-server-text
Retrieve one public asset by its exact kind and id. Source is returned only where the asset's contentLevel and license permit it. Remote templates include their actual file set only when the live source is available; if it is down, the tool returns a structured unavailable error with safe metadata and does not treat a download URL as source code. External and restricted assets expose metadata, license, attribution and source links only.

Args:
  - kind: the namespace returned by stylekit_list_assets.
  - id: the exact identifier within that namespace.

Returns complete structured JSON and JSON text, preserving the namespace and provenance.
```

### `stylekit_get_style`

- The name "stylekit_get_style" says "get": a word for reading.
- Tag untrusted-input: The server marks it openWorldHint: true: it reaches things outside the server's own data.

Description, as the server wrote it:

```untrusted-server-text
Get one style's profile and, optionally, its design tokens, one component recipe, and the full implementation brief. Everything ships through this single tool so the server stays small: pass include: ["tokens"], ["recipe"], ["brief"] or a combination.

Style data is read from the live catalogue with a five-minute cache; when live data is unavailable, a bundled snapshot may be returned and marked with origin and fallbackReason.

Args:
  - slug (string): style identifier, e.g. "glassmorphism", "neo-brutalist".
  - include (array, optional): sections to attach — "tokens", "recipe", "brief".
  - component (string, optional): recipe id for include: ["recipe"], e.g. "button"; defaults to the style's first recipe.

Returns structured JSON with the shape below. When a canonical design.md is available, the human-readable text includes its complete Markdown; the full machine-readable brief remains at its URL.

Examples:
  - "how should I use neo-brutalist?" -> slug: "neo-brutalist"
  - "give me the glassmorphism button code" -> slug: "glassmorphism", include: ["recipe"], component: "button"
  - Returns an error with a hint if the slug is unknown.
```

### `stylekit_list_assets`

- The name "stylekit_list_assets" says "list": a word for reading.
- Tag untrusted-input: The server marks it openWorldHint: true: it reaches things outside the server's own data. The description's first sentence says "browse": it reads web pages, which anyone can write.

Description, as the server wrote it:

```untrusted-server-text
Browse StyleKit's public asset catalogue by namespace and query. Use the returned kind/id pair with stylekit_get_asset.

Args:
  - kind (optional): one public namespace such as animation, component-pattern, template, or experience-pack.
  - query (optional): search public names, descriptions and tags.
  - offset (number >=0, default 0): results to skip.
  - limit (number 1-100, default 20): page size.

Returns the full JSON page, including availability, contentLevel, license, attribution, source provenance, and kind counts. An empty page is a successful result. Private user kits and unpublished submissions are not listed.
```

### `stylekit_search_styles`

- The name "stylekit_search_styles" says "search": a word for reading.
- Tag untrusted-input: The server marks it openWorldHint: true: it reaches things outside the server's own data.

Description, as the server wrote it:

```untrusted-server-text
Search StyleKit design styles by keyword and/or category, with pagination. Exact slug or name matches are ranked first.

Args:
  - query (string, optional): matches slug, name, description, tags, keywords (case-insensitive).
  - category ('modern'|'retro'|'minimal'|'expressive', optional): restrict to one category.
  - limit (number 1-50, default 15): page size.
  - offset (number >=0, default 0): results to skip (for paging).

Returns structured JSON with total, count, offset, has_more, source, optional fallbackReason, and results. A valid query with no matches returns an empty results array, not a tool error.

Examples:
  - "find a glassy frosted style" -> query: "glass"
  - "next page of retro styles" -> category: "retro", offset: 15
  - For full tokens of one style, call stylekit_get_style with include: ["tokens"] instead.
```

## Claude Code permissions

For `.claude/settings.json`, assuming the server is installed as `stylekit`:

```json
{
  "permissions": {
    "allow": [
      "mcp__stylekit__stylekit_search_styles",
      "mcp__stylekit__stylekit_list_assets",
      "mcp__stylekit__stylekit_get_asset",
      "mcp__stylekit__stylekit_get_style"
    ],
    "ask": [
      "mcp__stylekit__stylekit_lint_code"
    ],
    "deny": []
  }
}
```

Capability tags behind the rules (a read tagged outbound is asked about):

- `mcp__stylekit__stylekit_search_styles`: untrusted-input
- `mcp__stylekit__stylekit_list_assets`: untrusted-input
- `mcp__stylekit__stylekit_get_asset`: untrusted-input
- `mcp__stylekit__stylekit_get_style`: untrusted-input
- `mcp__stylekit__stylekit_lint_code`: untrusted-input

## Remote posture

What an anonymous client saw at the remote endpoint, last 2026-11-20T18:56:41.813Z, unchanged since 2026-11-13T11:55:10.876Z. No credentials were sent and no tool was called. Values are shown as the endpoint returned them. Method: https://radar.anlyon.com/methodology#remote-posture

| Fact | Observed |
| --- | --- |
| Reachable | `yes` |
| HTTP status of an anonymous initialize | `200` |
| tools/list served without credentials | `yes` |
| Tools that change things, listed without credentials | `1` |
| Protocol version | `2025-06-18` |
| WWW-Authenticate on 401 | `not observed` |
| resource_metadata in WWW-Authenticate | `none` |
| RFC 9728 protected resource metadata | `none` |
| Authorization servers | `none` |
| Legacy HTTP+SSE transport | `no` |
| Dynamic client registration | `not observed` |
| Client ID metadata documents | `not observed` |

- **warn** Tools that change things, listed to anyone ([unauth-remote-writes](https://radar.anlyon.com/methodology#unauth-remote-writes))

```untrusted-server-text
The endpoint served tools/list without credentials, and the list has 1 tool that changes things (1 write). Whether a call also works without credentials was not tested: tools are never called.

https://www.stylekit.top/mcp: tools/list returned 5 tools to an anonymous client, including stylekit_lint_code (write)
```

Posture changes:

- 2026-11-13T11:55:10.876Z: Tools that change things, listed without credentials `2` -> `1`
