Fake Claude Code Guides Lure Vibe Coders With Malware

Developers trying out Claude Code are being funneled to booby-trapped install guides that quietly deliver malware, according to new research from Push Security. The campaign, labeled InstallFix, clones legitimate documentation and swaps in a single malicious download URL, betting that “copy and paste” muscle memory will do the rest. Because the user launches the command, many endpoint defenses treat it as trusted activity, letting the payload slip through.

How the InstallFix Trap Works to Trick Developers

Attackers mirror the look and feel of official Claude Code pages—logos, layout, even working links—then alter the install snippet to point at attacker-controlled infrastructure. The familiar curl-to-bash style one-liner fetches a first-stage executable, which immediately pulls a second-stage bundle from a remote server. Push Security notes the technique is a twist on the broader ClickFix social engineering pattern, but here the “fix” is an installation step users are already primed to trust.

This approach capitalizes on a well-known weak spot in developer workflows: pasting shell commands from the web into a terminal without verifying the source or inspecting the script. In MITRE ATT&CK terms, it’s classic User Execution (T1204), with additional defense evasion as the chain masquerades as a legitimate installer. Once the script runs, traditional antivirus can be late to the party.

Why Vibe Coders Are Being Targeted by Attackers

“Vibe coding” lowers the barrier to software tinkering, and that’s a double-edged sword. Newcomers experimenting with AI assistants are more likely to follow quick-start guides verbatim, especially when documentation promises a one-command setup. Industry surveys from GitHub and Stack Overflow show strong, rising interest in AI coding helpers, which widens the pool of potential victims who are comfortable pasting commands but less comfortable auditing them.

Malvertising Supercharges the Lure for Claude Code

Push Security observed malicious guides appearing as sponsored results when users searched for Claude Code, a tactic that can place the trap above the legitimate site. Search malvertising isn’t new, but it’s effective: even seasoned developers often click the first result if it looks official. Google’s Ads Safety Report regularly cites the removal of more than 5 billion bad ads annually, yet adversaries still slip through with short-lived domains and convincing brand mimicry.

What the Malware Steals from Compromised Systems

The second-stage payload in this campaign appears linked to the Amatera Stealer family, targeting high-value data like browser passwords, cookies, and session tokens. Token theft is particularly damaging because it can enable account access even when users have strong passwords and MFA. CISA and multiple incident reports have warned that session hijacking can persist until tokens are revoked and all devices are fully sanitized.

Researchers also note that these stealers often include update logic or persistence hooks, making simple removal unreliable. If you executed a suspicious install command, assume exfiltration occurred and rotate credentials, invalidate tokens, and reimage high-risk systems as needed.

How to Install Safely Right Now Without Risky Shortcuts

Verify the domain before you copy anything. Check for subtle typos, extra words, or unusual TLDs, and prefer bookmarks to known-official docs. If a guide asks you to paste a curl-to-bash command, open the URL in a browser first, inspect the script’s contents, and confirm it lives under the project’s official domain or repository.

Use trusted package managers over ad hoc scripts when possible. Prefer signed releases and verify checksums from the same official source; mismatches are a red flag. Avoid shortened links in install steps, and be skeptical of “one-line” installers that request elevated privileges without clear justification.

Harden the blast radius: run new tools in containers or disposable VMs, reduce persistent browser sessions, and store tokens in dedicated profiles with limited scope. Enforce phishing-resistant MFA (such as FIDO2 security keys), which diminishes the value of stolen credentials and weakens session hijacking attempts.

What Security Teams Should Do to Counter InstallFix

Block newly registered and typosquatted domains at DNS where possible, and monitor egress for paste-and-run patterns pulling from non-official hosts. Educate developers on recognizing malvertising, and steer them to a single, vetted landing page for tools like Claude Code. Detection content focused on suspicious shell one-liners and unsigned installers can catch misuse without breaking legitimate builds.

InstallFix is likely to proliferate beyond Claude Code to any fast-growing AI tool with a “quick start” audience. The fix isn’t to abandon speed, but to make speed safer: verify sources, inspect scripts, and never let convenience outrun scrutiny.