cloud-gcp
Warn
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command (
ls) to inspect the contents of multiple local directories, including hidden folders associated with the AI platform (~/.claude/skills/,~/.agent/skills/, etc.). This constitutes environment fingerprinting to identify other installed skills. - [EXTERNAL_DOWNLOADS]: The skill's instructions and documentation recommend downloading and installing companion skills from an external GitHub repository via the npm registry.
- [REMOTE_CODE_EXECUTION]: The recommendation to use
npx skills addfacilitates the download and immediate execution of remote code on the user's host machine.
Audit Metadata