Add comprehensive analysis of client information collection for risk control and Pro/Max subscription users - #1

Analyzed Claude Code v2.1.88 codebase to document all client information collected for risk control, user identification, and how subscription tiers affect data collection and feature access.

Analysis Reports

1. General Client Information Collection (客户端信息收集分析报告.md)

Comprehensive documentation of all data collection mechanisms across 15 categories:

• User Identification: 32-byte persistent deviceId, sessionId, email from OAuth
• Environment Detection: 30+ IDEs/terminals (VSCode, JetBrains, Cursor), cloud platforms (AWS, GCP, Azure), CI/CD systems (GitHub Actions, GitLab CI)
• System Metrics: Memory (RSS, heap), CPU usage, process uptime, package managers, runtimes
• Analytics Infrastructure: Anthropic API (/api/event_logging/batch), Datadog integration with 64 event types
• Privacy Controls: SHA256 user bucketing (30 buckets), PII tagging, feature-gated collection, dynamic event sampling

2. Pro/Max Subscription Analysis (Pro-Max订阅用户数据收集分析.md)

Detailed breakdown of subscription-specific data collection and feature gating:

Pro/Max Additional Data Collection:

type CoreUserData = {
  subscriptionType?: 'pro' | 'max'    // Subscription tier
  rateLimitTier?: string               // Rate limit tier (e.g., default_claude_max_20x)
  billingType?: BillingType            // stripe_subscription | apple_subscription | ...
  accountCreatedAt?: string            // For DAU-by-age analysis
  subscriptionCreatedAt?: string       // For churn analysis
  hasExtraUsageEnabled?: boolean       // Overage monitoring
}

Subscription-Gated Features:

• Max-only: Guest Pass feature (src/services/api/referral.ts:71-77)
• Pro: Medium default effort level, automatic billing access
• Max: High default effort level, 20x rate limits, all Pro features

Code References: All findings include specific file paths and line numbers for verification (e.g., src/utils/auth.ts:1564-1730 for subscription type detection).

Data Flow

Client → Local Queue (~/.claude/telemetry/)
       → Batch Export (10s/200 events)
       → [Anthropic API | Datadog]

Both reports provide actionable intelligence for understanding risk control mechanisms, privacy protections, and subscription tier differentiation in the codebase.