gangtise-copilot

SUSPICIOUS: the skill's purpose and credential use are largely coherent, and auth traffic appears to go to Gangtise directly, but its footprint is still risky because it installs 19 additional skills from mutable, unpinned ZIP bundles and extends trust to upstream content the user cannot easily verify. This looks more like a high supply-chain/transitive-install risk than confirmed malware or credential theft.

The fragment primarily implements a remote skill installer and does not itself show clear malware such as credential theft, exfiltration, or persistence beyond intended agent skill installation. However, it downloads and extracts unauthenticated archives, installs their contents into agent search paths, and uses unvalidated skill names in destructive filesystem operations. These are significant supply-chain and path-handling risks that should be remediated with pinned HTTPS URLs, cryptographic verification, archive path/symlink validation, strict skill-name validation, and secure temporary directories. Because the file is truncated and key variable definitions are unavailable, confidence is reduced.

The fragment implements a plausible Gangtise credential setup and verification workflow rather than clearly malicious behavior. It handles highly sensitive credentials, transmits them to an endpoint not visible here, stores them in plaintext with restrictive permissions, and propagates them through symlinks. The endpoint and preceding variable definitions must be verified before use. The response-to-Python interpolation, unescaped JSON construction, terminal response printing, and temporary credential backups present security weaknesses. No clear malware, unrelated data theft, persistence, or destructive action is demonstrated in the supplied portion.