anti-claude-checklisted

# 反 Claude 检查 Audit privacy leaks, contradictory network signals, and documented Claude Code privacy controls without trying to defeat platform safeguards. Prefer stable, ordinary system behavior and the smallest defensible configuration change. ## Safety Boundary - Work read-only by default. Ask before changing proxy, DNS, browser, or operating-system settings. - Never elevate silently. Before an administrator-required change, show the exact scope, obtain approval, and ask the user to relaunch the controlling app or terminal as Administrator. - Do not help spoof identity or browser fingerprints, bypass bot or fraud controls, evade regional restrictions, or support multi-account abuse. - Do not recommend anti-detect browsers for risk-control evasion. - Do not collect or expose cookies, passwords, subscription URLs, proxy credentials, payment details, or fabricated identity and billing data. - Redact usernames, node names, IP addresses, and other sensitive identifiers before publishing reports. ## Triage External Tools Do not install, execute, or copy a third-party repository merely because it appears in an audit. Inspect its stated purpose first, then classify it: - Exclude [Camoufox](https://github.com/daijro/camoufox), [CloakBrowser](https://github.com/CloakHQ/cloakbrowser), and [browser-profiles](https://github.com/aitofy-dev/browser-profiles) from the workflow. Their documented features include anti-detect behavior, automation concealment, or fingerprint spoofing. N