What caused the Anthropic Claude Code source code leak?

A source map file in npm package v2.1.88 linked to an unprotected TypeScript ZIP on their R2 bucket — human packaging error with Bun runtime defaults.

What unreleased features were in the leaked Claude Code code?

KAIROS daemon, ULTRAPLAN cloud planning, BUDDY AI pet, plus codenames like Capybara for Claude 4.6 and Undercover Mode to hide internals.

Is the Claude Code leak a major security risk?

No customer data lost, but exposes agent logic, prompts, and defenses — aiding reverse-engineering, though minified code was somewhat accessible before.

💻 Programming Languages

Anthropic's Claude Code Leak: Source Maps Spill AI Agent Guts Wide Open

Anthropic promised a smoothly AI coding sidekick. Then a tiny npm packaging flub dumped its entire source code online, revealing a beast of an agent system lurking beneath.

Open Source Beat Apr 07, 2026 4 min read

Screenshot of leaked Claude Code source map file from npm package exposing TypeScript codebase

⚡ Key Takeaways

Accidental npm source map exposed 512K+ lines of Claude Code TypeScript, revealing complex agent architecture. 𝕏
Unreleased features like KAIROS daemon and BUDDY AI pet hint at Anthropic's wild roadmap. 𝕏
Echoes Netscape leak; likely sparks AI agent innovation wave despite 'no breach' spin. 𝕏

Published by

Open Source Beat

Community-driven. Code-first.

#AI agent architecture #Anthropic Claude Code #Anthropic leak #Claude Code source #ai-agents #npm security #source code leak

Worth sharing?

Get the best Open Source stories of the week in your inbox — no noise, no spam.

Originally reported by InfoQ

⚡ Key Takeaways

The 60-Second TL;DR

Open Source Beat

Share this article

Worth sharing?

Related Stories

Kubernetes' New LLM Stalker: OpenLIT Operator's Zero-Code Snooping

Veridian Guard: One Decorator to Save Your AI Agents from LLM Chaos

Anthropic's One-Line Fumble Leaks Billions in Code

Warden v2.0: Free CLI That Sniffs Out Malicious npm Packages in Seconds

Stay in the loop