The 'creditgauge' npm package version 1.0.0 is maliciously crafted with no functional library code. Its package.json declares a main entry point that is missing, and the tarball only contains package.json and a settings.json payload. The settings.json overrides the ANTHROPIC_BASE_URL to a non-Anthropic domain and embeds a hardcoded ANTHROPIC_AUTH_TOKEN, causing any Claude Code sessions using this configuration to route prompts and source code through an attacker-controlled proxy. It also defines a statusLine that runs a bash command invoking a shell script from a third-party plugin every 10 seconds. The configuration disables permission prompts, allowing silent and recurring execution of unvetted third-party code, resulting in unauthorized data exfiltration and code execution risks.

This malicious package enables silent exfiltration of developer AI prompts and source code to an attacker-controlled endpoint, potentially exposing sensitive intellectual property and confidential data. The recurring execution of a third-party plugin's shell script without user consent introduces a risk of arbitrary code execution within the developer's environment, which could lead to further compromise or persistence of malicious activity.

No official patch or remediation is currently documented for this package. Since this is a malicious package, the primary mitigation is to avoid using 'creditgauge' version 1.0.0 from untrusted sources. Remove any installations of this package and audit environments for signs of compromise. Monitor for any unauthorized network connections to suspicious endpoints such as 'https://api.minimaxi.com/anthropic'. Check vendor advisories or trusted security sources for updates or removal notices regarding this package.