Core Insights - Anthropic accidentally exposed the complete TypeScript source code of Claude Code by failing to exclude the .map files during npm package release, leading to significant community engagement with over 500,000 views and 2,300 likes within hours [2][3]. Group 1: Code Leak Details - The source map file (cli.js.map) was 57MB in size and allowed anyone to reverse-engineer the original source code using common tools, highlighting a significant oversight in package management [5][7]. - The leak was not a sophisticated attack; it was a simple mistake that could be replicated by anyone familiar with npm [7][6]. Group 2: Hidden Features Discovered - The community uncovered at least five hidden features within the leaked code, including KAIROS, a mode that allows Claude Code to run continuously in the background, functioning as an always-on agent [8][10]. - Other features included a complete electronic pet system called Buddy, which had various species and attributes, indicating a playful aspect within the code [12][13]. - The "Undercover Mode" feature raised ethical concerns as it allows the system to hide AI-generated contributions in open-source projects, contradicting the industry's push for transparency [14][16]. Group 3: Architectural Insights - The leaked directory structure revealed a comprehensive AI agent architecture, including modules for coordination, tools, context management, and a semantic search engine, indicating a sophisticated design for multi-agent collaboration [22][19]. - The architecture suggests that Anthropic is developing a fully autonomous development system capable of operating 24/7 without human intervention [26][19]. Group 4: Industry Implications - The incident raises questions about the transparency of AI-generated code, especially as Anthropic's tools actively conceal AI involvement, which may conflict with industry standards [26][16]. - Companies are advised to review their npm packages to prevent similar leaks, emphasizing the importance of rigorous package management practices [26][26]. - The playful inclusion of an electronic pet in the code serves as a reminder of the human element behind AI development, contrasting with the serious discussions surrounding AI risks [26][26].
Claude Code 源码泄露:5 个隐藏功能曝光
深思SenseAI·2026-03-31 10:08