Show HN: Apple-design-hig agent skill, DocC updater CLI, and MCP server A developer released apple-design-hig, an open-source toolkit that queries Apple's live, unauthenticated DocC JSON endpoints and compiles over 172 official Human Interface Guidelines topics into deterministic GitHub-Flavored Markdown for AI coding agents. The project ships three synchronized surfaces: a bundled agent skill with a zero-dependency standard-library fetcher (scripts/hig_fetch.py), a FastMCP server over stdio named apple-design-hig-mcp exposing four tools (hig_list_sections, hig_search, hig_get_topic, hig_sync) to hosts including Claude Desktop, Cursor, Windsurf and Zed, and a uv-managed CLI for searching the live HIG catalog and syncing offline references. The toolkit targets the problem of AI coding assistants relying on stale training weights and guessing touch-target dimensions or hallucinating deprecated navigation bars when generating Apple-platform views. Authoritative, deterministic access to Apple's Human Interface Guidelines for AI coding agents, design systems engineers, and Model Context Protocol MCP clients. Overview overview | Architecture architecture | Quick Start quick-start | MCP Server model-context-protocol-mcp-server | CLI Reference command-line-interface-cli | Agent Skill bundled-agent-skill | Platform Matrix design-standards--platform-matrix | CI Sync automated-upstream-synchronization | Development development--verification Interfaces demand precision. Yet most AI coding assistants rely on stale training weights when generating or auditing Apple-platform views - guessing touch-target dimensions, hallucinating deprecated navigation bars, or ignoring visionOS eye-gaze ergonomics and Liquid Glass material semantics. apple-design-hig replaces guesswork with primary-source telemetry. It queries Apple's live, unauthenticated DocC JSON endpoints directly, traverses the canonical abstract syntax tree AST for over 172 official Human Interface Guidelines topics , and compiles them into clean, deterministic GitHub-Flavored Markdown. Engineered for zero-friction adoption across local terminals and sandboxed agent runtimes, this repository unifies three synchronized delivery surfaces: 1. Bundled Agent Skill skills/apple-design-hig/ https://github.com/mr8lu/apple-design-hig/blob/main/skills/apple-design-hig : A portable, self-contained skill package featuring Apple's six core design principles, Section 508 / WCAG AA-AAA contrast mandates, pre-rendered Markdown references across all six HIG pillars, and a zero-dependency standard-library fetcher scripts/hig fetch.py https://github.com/mr8lu/apple-design-hig/blob/main/skills/apple-design-hig/scripts/hig fetch.py . 2. Model Context Protocol Server apple-design-hig-mcp : A FastMCP server communicating over stdio that exposes four structured tools hig list sections , hig search , hig get topic , hig sync to Gemini CLI, Claude Desktop, Cursor, Windsurf, Zed, and any MCP-compliant host. 3. Command-Line Interface apple-design-hig : A fast terminal toolkit managed via uv to search the live HIG catalog, stream full topic specifications as Markdown, synchronize offline reference files, and provision local agent skill directories. Every query and synchronization pass flows from Apple's production DocC JSON endpoints through a deterministic AST resolver and ASCII-normalized Markdown compiler. flowchart LR subgraph Apple "Apple Developer DocC JSON API" IDX "Index Endpoint