MalwarePublished 26 May 2026Updated 15 Jul 2026

ggk-happy (npm) · Malicious Code

Affects ggk-happySource GCVE Database

Malicious package : remove ggk-happy and rotate any secrets on hosts that installed it.

Data as of 15 Jul 2026, 12:36 UTC

Classification

Malicious package

Published to npm · OSV MAL-2026-4789

What to do

Remove ggk-happy from every project, lockfile and image that pulls it in. Treat any host that installed it as compromised: rotate the credentials, tokens and keys it could read, from a clean machine.

Affected software

Product	Affected	Fixed in
npmggk-happy	Affected 1.0.9 – 1.2.34 11 affected builds	Fixed in–

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

Description

--- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (d747cc7acd8bbc1defbf475aa3202fd332fd43135a6fedddfcc37356ce02eb54) Package presents itself as the legitimate slopus/happy CLI (README instructs `npm install -g happy`; homepage, repository, bugs, and author fields all point at happy.engineering / github.com/slopus/happy), but the published name is ggk-happy and the shipped code has been modified in ways that harm installers: 1. Backend redirection: DEFAULT_SERVER_URL is hardcoded to https://happy-api.ask-ggk.com and DEFAULT_WEBAPP_URL to https://happy.ask-ggk.com — third-party hosts not disclosed in the README, which still references api.cluster-fluster.com. All coding-agent sessions (Claude Code / Codex prompts, source code, auth tokens, machine metadata) flow to ask-ggk.com instead of the publisher advertised in package metadata. 2. Remote-control surface pointed at attacker infrastructure: ApiSessionClient / ApiMachineClient open a socket.io connection to the redirected server and register RPC handlers named `bash`, `readFile`, `writeFile`, `listDirectory`, `getDirectoryTree`, `ripgrep`, `difftastic`. The `bash` handler runs `execAsync(data.command, options)` on parameters supplied by the remote server. Whoever operates happy-api.ask-ggk.com can execute arbitrary shell commands and read/write files on the installer's machine. 3. Install-time auto-takeover: scripts/postinstall.cjs, on global or root install, unconditionally spawns `node bin/happy.mjs install`, running the full auth/registration flow (browser opened to happy.ask-ggk.com, machine registered to happy-api.ask-ggk.com, daemon started) at `npm install -g` time. Auto-takeover is opt-out via env var, not opt-in, so `sudo npm install -g ggk-happy` immediately binds the machine to the third-party backend, with elevated privileges on Linux/macOS root installs. 4. Unpinned binary dropper: RTK_DOWNLOAD_URLS fetches platform-specific `rtk` binaries from https://minio.ask-ggk.com/happy/rtk-*.{zip,tar.gz} (win32/darwin/linux, x64/arm64) with no hash or signature verification, stages them to ~/.local/share/ggkhappy/rtk/bin/rtk (or %LOCALAPPDATA%\ggkhappy\rtk\bin\rtk.exe), and later invokes them via execFile. Purpose of the binary is not documented. 5. Covert identity reporting: readRtkIdentity() reads the installer's machineId from ~/.happy settings and the auth token from ~/.happy/access.key and POSTs them, together with platform/version fields, to https://guguke.ask-ggk.com/api/v1/agent/rtk/gain/report — a side channel to a host not referenced in README documentation. This is not a legitimate fork: the surface identity (name, homepage, repo, README install command) still claims to be slopus/happy, while all backends, the remote-shell RPC destination, the runtime binary source, and the identity reporter have been repointed to ask-ggk.com infrastructure. Installers who follow the README believe they are installing the upstream project and instead grant an undisclosed operator remote code execution over their machine and full visibility into their AI coding sessions.

Impact

Attackers controlling the backend infrastructure can execute arbitrary shell commands with the privileges of the installing user, including root on Linux/macOS if installed globally. They gain full read/write access to the victim's file system and can monitor or manipulate AI coding sessions and authentication tokens. The automatic installation-time takeover means victims are compromised immediately upon installing the package globally. The unsigned binary dropper further expands the attack surface with unknown payloads. Overall, this leads to complete compromise of the victim's machine and sensitive data exposure.

Defensive guidance

Users should immediately avoid installing or using the 'ggk-happy' package. Verify package authenticity by checking the exact package name and source repository before installation. Remove any installations of 'ggk-happy' from affected systems and investigate for signs of compromise. Since this is a malicious package impersonating a legitimate tool, rely only on the official 'happy' package from the verified source. Monitor for updates from the legitimate project or npm advisories for further guidance.

Summary

The 'ggk-happy' npm package impersonates the legitimate 'happy' CLI by retaining its name, homepage, and repository metadata while redirecting all backend API and webapp URLs to attacker-controlled domains (happy-api.ask-ggk.com, happy.ask-ggk.com). It establishes a socket.io connection exposing RPC handlers that allow arbitrary shell command execution and file system manipulation. During global or root installation, it automatically initiates an authentication and registration flow binding the machine to the attacker's backend with elevated privileges. The package downloads platform-specific 'rtk' binaries from an unverified source and executes them without integrity checks. Additionally, it exfiltrates machine identifiers and authentication tokens to a covert reporting server (guguke.ask-ggk.com). This is a supply chain attack delivering remote code execution and data exfiltration under the guise of a trusted package.

References

Community reviews

Loading reviews…

What to do

Affected software

Description

AI analysisMachine-generated, updated 11 Jul 2026, 09:58 UTC

Technical details

References

Community reviews