skill-library

SUSPICIOUS: the skill’s behavior largely matches its stated purpose as a package-library installer, but that purpose itself is high-trust. Its main risk is transitive installation of remote agent packages from a personal GitHub repository without integrity verification or pinning, plus the ability to install executable utilities. I found no direct credential harvesting, covert exfiltration, or malicious pre-execution behavior in the provided skill text.

The procedure appears intended for legitimate skill updates and contains no clear malware payload. It presents a meaningful supply-chain risk because remote, mutable skill content is downloaded and overwrites local copies without commit pinning, signatures, checksums, or content validation. Unquoted skill-name interpolation and the shared /tmp cache add potential path-manipulation and tampering risks. Use only with a trusted repository and add strict input validation and integrity verification.

The code is an installer workflow rather than overt malware. It performs network retrieval and writes remote skill content into a user configuration directory, creating a meaningful software supply-chain risk due to absent signature/checksum verification, mutable remote references, redirect following, and insufficient input validation. No direct credential theft, exfiltration, reverse shell, or destructive payload is present in the fragment.

The code describes a legitimate uninstall operation, but its shell-command construction is unsafe unless <name> is strictly validated and safely quoted. The primary security concern is destructive path manipulation or command injection through an untrusted skill name. No direct malicious behavior is evident in the provided procedure.