On Oct 19 (21:47 UTC) GitHub published malware advisories for eight npm packages that impersonate setup or activation tools for popular AI assistants: chatgpt-pro-activator (GHSA-qp43-mgx6-56x2), chatgpt-pro-setup-tool (GHSA-xwx3-5h3q-2v8r), chatgpt-desktop-setup (GHSA-pgq5-xq28-533m), @decoopus/chatgpt-setup (GHSA-4632-2g34-4r9w), claude-desktop-setup (GHSA-f4mf-vhm8-xxqc), @decoopus/claude-setup (GHSA-p7p5-jrq5-q47x), codex-setup (GHSA-h8cg-xwfv-9x77) and api-anthropic (GHSA-cj9w-7h22-q93g). GitHub's standard malware wording applies: any computer with the package installed or running "should be considered fully compromised". Rotate all secrets from a different computer, and know that removing the package may not remove what it installed. The advisories don't describe the payload.

None of these are official. The official SDKs are openai and @anthropic-ai/sdk, and the desktop apps don't install from npm. The look-alike api-anthropic name is the kind a coding agent or a hurried developer could easily type or accept. Do now: search lockfiles, CI images, dev containers and agent sandboxes for these eight names (for example npm ls or a grep across package-lock.json/pnpm-lock.yaml). If any turns up, isolate the machine, rotate every key it held, and rebuild. For agents that can run npm install, require approval for new dependencies and use an allowlist or registry proxy.