HIG Doctor

Registre officiel Outils de développement

Apple Human Interface Guidelines search, lookup, and project audits for AI coding agents.

Installation

Ajoutez ceci à la configuration MCP de votre client (Claude Desktop : claude_desktop_config.json ; Cursor : .cursor/mcp.json ; VS Code et les autres utilisent la même forme « mcpServers »).

NPM package hig-mcp
Claude Code
claude mcp add hig-doctor -e 'HIG_SKILLS_DIR=<HIG_SKILLS_DIR>' -- npx -y hig-mcp@0.2.3
Configuration JSON (Claude Desktop, Cursor, VS Code)
{
    "mcpServers": {
        "hig-doctor": {
            "command": "npx",
            "args": [
                "-y",
                "hig-mcp@0.2.3"
            ],
            "env": {
                "HIG_SKILLS_DIR": "<HIG_SKILLS_DIR>"
            }
        }
    }
}

Variables d'environnement

  • HIG_SKILLS_DIR Optional override for the skills directory. The published package bundles the HIG skills corpus, so this is only needed to point the server at a different or newer local copy.

Documentation

Tirée du README du projet sur GitHub, actualisée chaque semaine. Les liens ouvrent GitHub.

Sommaire
  1. HIG Doctor
  2. Run an audit
  3. Why use HIG Doctor
  4. Choose a surface
  5. How analysis works
  6. Framework coverage
  7. Skills corpus
  8. Limits and evidence boundary
  9. Documentation
  10. Raintree open-source system
  11. Project policies

HIG Doctor

Active open-source project · MIT tooling with attributed Apple reference content

HIG Doctor helps developers and coding agents find interface issues before release. It checks Apple-platform source against Apple’s Human Interface Guidelines and checks web and cross-platform source against aligned accessibility and interface-quality rules.

Run an audit

npx hig-doctor .

The CLI detects the project frameworks, reports concerns by severity, and points to a specific fix and source reference. This result is generated from the committed test/fixtures/readme-audit project and checked in the test suite:

2 moderate concerns · swiftui · 1 file
View.swift:5 · swift/navigation-view-deprecated
View.swift:7 · swift/hardcoded-color

The audit catalog currently contains 431 rules across 14 frameworks. Counts are generated from the rule catalog and checked in CI; they are coverage inventory, not a claim of complete HIG conformance.

Why use HIG Doctor

  • Catch reviewable source issues. Findings include severity, location, rule ID, fix guidance, and the reference that supports the concern.
  • Gate only new debt. Configuration, inline suppressions, content-based baselines, SARIF, and --fail-on support gradual adoption.
  • Give agents bounded guidance. MCP tools search the frozen reference corpus and explain individual findings without presenting generated advice as canonical HIG.
  • Use one engine across workflows. The CLI, MCP server, and embeddable core package share the same catalog and analysis tiers.

Choose a surface

SurfaceUse it forStart
Audit CLISource scans and CI gatesnpx hig-doctor .
MCP serverSearch, lookup, file audits, and explanationsnpx -y hig-mcp
Agent skillsDesign guidance and source auditsCanonical agent skills
Core packageEmbed the rule enginenpm install @raintree-technology/hig-doctor-core

The MCP server works over stdio or streamable HTTP. Its six tools list skills, look up topics, search the corpus, audit projects or files, and explain findings. See the MCP package README for client configuration.

How analysis works

framework detection → regex scan → structural refinement → categorized findings → report or SARIF
  • The zero-dependency regex tier is comment- and string-aware.
  • Swift structural analysis follows chained modifiers to remove handled findings.
  • The TypeScript compiler refines selected JSX accessibility checks when available.
  • Every finding records the engine that produced it.

Precision and recall are measured on an annotated fixture corpus in docs/benchmark.md. CI enforces the published floors.

Framework coverage

Apple-platform rows are checked against the HIG directly. Web and cross-platform rows use universal accessibility and interface-quality principles that align with the HIG.

FrameworkRulesBasis
SwiftUI (swift)70Apple HIG
UIKit35Apple HIG
AppKit25Apple HIG
watchOS13Apple HIG
visionOS10Apple HIG
React / Next.js (web)122HIG-aligned universal checks
CSS / SCSS25HIG-aligned universal checks
Vue / Nuxt19HIG-aligned universal checks
Angular17HIG-aligned universal checks
Svelte / SvelteKit14HIG-aligned universal checks
Flutter21HIG-aligned universal checks
Jetpack Compose28HIG-aligned universal checks
Android XML18HIG-aligned universal checks
React Native14HIG-aligned universal checks

The authoritative per-rule inventory is docs/rules.md.

Skills corpus

The frozen 2025-02-02 snapshot contains 14 skills and 156 reference topics. Apple’s live Human Interface Guidelines remain canonical.

SkillDescription
hig-foundationsColor, typography, symbols, accessibility, layout, motion, privacy, and branding
hig-platformsiOS, iPadOS, macOS, tvOS, watchOS, and visionOS
hig-patternsOnboarding, navigation, search, feedback, modality, and settings
hig-inputsTouch, Pencil, keyboard, controller, pointer, Crown, and eye input
hig-technologiesSiri, Apple Pay, HealthKit, ARKit, ML, Sign in with Apple, and SharePlay
hig-project-contextShared project context for tailored guidance
hig-components-contentCharts, collections, images, web views, and lockups
hig-components-controlsPickers, toggles, sliders, buttons, and fields
hig-components-dialogsAlerts, action sheets, popovers, and sheets
hig-components-layoutSidebars, split views, tabs, lists, tables, and windows
hig-components-menusMenus, toolbars, buttons, and menu bars
hig-components-searchSearch fields, page controls, and path controls
hig-components-statusProgress indicators, status bars, and activity rings
hig-components-systemWidgets, live activities, notifications, complications, and App Clips

Nightly drift detection compares the snapshot with Apple’s DocC JSON. Content changes remain human-reviewed; a hash change does not automatically rewrite guidance.

Limits and evidence boundary

Automated findings support review. They do not prove accessibility, HIG conformance, or design quality. Regex fallback can produce different precision than structural analysis, and project-specific context can justify a documented suppression.

Apple owns the HIG content. This repository provides organization, cross-referencing, and detection rules. Each reference retains attribution and a canonical source URL.

Documentation

Raintree open-source system

HIG Doctor owns interface guidance and source audits. It can be used independently. DocPull acquires evidence, PolicyStrata tests policy behavior, Trellis enforces shared code policy, and Raintree Standards defines governed requirements. See the Raintree open-source portfolio.

Raintree also builds BetterOff, which keeps a household's spending, investments, assets, debts, and net worth in one current record.

Project policies

Contributing · Code of Conduct · Security · Changelog · Source repository · MIT License · Third-party notices

Apple HIG reference text in skills/*/references/ is © Apple Inc. and remains subject to Apple’s terms.

Questions fréquentes

Qu'est-ce que HIG Doctor ?

Apple Human Interface Guidelines search, lookup, and project audits for AI coding agents.

Comment installer HIG Doctor ?

Il s'exécute sur votre machine. Dans Claude Code : claude mcp add hig-doctor -e 'HIG_SKILLS_DIR=<HIG_SKILLS_DIR>' -- npx -y hig-mcp@0.2.3. Pour les autres clients, collez la configuration JSON de cette page dans leurs réglages MCP.

Faut-il une clé API pour HIG Doctor ?

Son entrée au registre n'indique aucune clé ni secret obligatoire. Certaines fonctions peuvent tout de même demander un compte auprès du service auquel il se connecte.

HIG Doctor est-il open source ?

Oui. Son code est sur GitHub sous licence MIT.

Qui maintient HIG Doctor ?

Il est publié par raintree-technology. AI.tn le référence sans le maintenir ni le recommander.

Issu du registre MCP officiel. Être listé n'est pas une recommandation : vérifiez le code et les permissions d'un serveur avant de lui donner accès à vos données.