On September 23, 2026, malicious versions of the MemoryOS Python package (version 2.0.34) were published containing a prebuilt Go implant called sckit. Upon import, the package launches a detached Go binary that collects credential files such as .npmrc, .pypirc, .git-credentials, SSH private keys, vault tokens, and various token caches from the user's home directory. The implant exfiltrates these credentials to multiple attacker-controlled domains under skyleen.fr using a secure protocol involving CBOR, X25519 key exchange, and XChaCha20-Poly1305 encryption. The implant also has code to copy itself into other repositories and packages accessible with stolen credentials. The package's build backend modifies CI environment variables to execute a signed payload that steals publishing tokens from GitHub Actions environments, enabling the attacker to compromise downstream package publishing. The implant deletes its binary after execution to evade detection. This is a supply-chain compromise with clear malicious intent targeting credential theft and propagation.

The implant steals a wide range of sensitive credentials from the user's home directory, including package manager credentials, Git credentials, SSH private keys, vault tokens, and cached tokens. These stolen credentials are sent to attacker-controlled infrastructure, enabling further compromise of developer environments and potentially other repositories and packages. The CI environment token theft allows attackers to hijack downstream package publishing workflows, increasing the scope of the supply-chain compromise. The implant's self-deletion and encrypted communications make detection and analysis more difficult.

No official patch or remediation is stated in the provided data. Users should immediately stop using version 2.0.34 of the MemoryOS package and remove it from their environments. Audit and revoke any potentially compromised credentials, including package manager tokens, SSH keys, and CI environment tokens. Review CI/CD pipelines for unauthorized environment variable modifications or payload executions. Monitor for suspicious network connections to the indicated attacker domains. Check the vendor advisory or PyPI for updates or fixed versions before resuming use.