취약점

-= Per source details. Do not edit below this line.=-

Source: amazon-inspector (8a08b3e13079279fb9dce40859dd868b0953bec139996eb7ac915a7dc415b29c)

Package is a third-party reconstruction of Anthropic's Claude Code CLI that misrepresents itself as the official product. package.json describes itself as 'Official Claude Code CLI — AI-powered coding assistant'. The bundled dist/main.js reuses Anthropic's production OAuth CLIENT_ID (9d1c250a-e61b-44d9-88ed-5944d1962f5e), the macOS keychain service name 'Claude Code', the MDM preference domain com.anthropic.claudecode, and the Windows policy registry path HKLM\SOFTWARE\Policies\ClaudeCode. At CLI startup it executes security find-generic-password -a <user> -w -s "Claude Code" to read OAuth tokens that the genuine @anthropic-ai/claude-code client stored under that identical keychain key. A user who installs this package believing it to be the official tool will have their existing Anthropic credentials read by an unaffiliated third-party binary, and any subsequent OAuth flow occurs under Anthropic's client identity without authorization. Although outbound traffic in the observed code paths goes to api.anthropic.com / platform.claude.com (no third-party exfiltration endpoint), the impersonation itself — combined with cross-vendor credential reuse — constitutes installer harm: the installer's trust in the Anthropic brand is exploited to grant a different vendor access to credentials the installer never intended to share with that vendor.