Package anthropic-claude-latest claims to be an 'Official Anthropic Claude SDK wrapper' but ships no Anthropic SDK code; the README is for an unrelated package cachesync-helper. On construction of the exported CacheSync / createCache, a _warmup routine schedules _prefetch after a 3-7 second random delay. _prefetch opens a TLS connection (with rejectUnauthorized: false) to one of four hardcoded IPs (104.194.134.33, 104.194.133.89, 107.189.20.82, 107.189.20.146:8443), receives base64-encoded files, and writes them to OS-camouflaged directories (~/Library/Application Support/com.apple.security, ~/.local/share/prometheus, %LOCALAPPDATA%\Microsoft\Windows Security\Health). The package then runs pip install --quiet --disable-pip-version-check for wallet/seed-phrase libraries (bip-utils, mnemonic, eth-account) and spawns python3 main.py detached with Object.assign({}, process.env, { _INTERNAL: '1' }) — handing the entire installer environment (AWS/GCP/NPM/SSH credentials, etc.) to the attacker-controlled payload. A .cs_v2 marker file is written for persistence gating. All sensitive identifiers (child_process, execFileSync, spawn, module.constructor._load, base64, python3, main.py) are split into [...].join('') arrays at lib/index.js to evade static scanners. This is a typosquat-lure remote code execution dropper targeting developers searching for Anthropic Claude SDKs.

Any computer that has this package installed or running should be considered fully compromised. All secrets and keys stored on that computer should be rotated immediately from a different computer. The package should be removed, but as full control of the computer may have been given to an outside entity, there is no guarantee that removing the package will remove all malicious software resulting from installing it.