modern-git
Fail
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill provides instructions to execute remote shell scripts using
curl | shandcurl | bashfor GitButler and Claude Code tools.\n- [REMOTE_CODE_EXECUTION]: Instructions involve downloading external binaries, modifying execution permissions withchmod +x, and running them locally.\n- [COMMAND_EXECUTION]: Suggests usingsudoto move binaries into system-wide protected directories like/usr/local/bin/.\n- [COMMAND_EXECUTION]: Performs persistent modifications to the user's environment by editing shell profiles (~/.bashrc,~/.zshrc) and global Git configurations.\n- [COMMAND_EXECUTION]: The commandgit maintenance startis used to create persistent background scheduled tasks on the system.\n- [EXTERNAL_DOWNLOADS]: Fetches scripts and binaries fromgitbutler.comandclaude.ai. Note thatclaude.ai/install.shis a non-standard installation path for the tool, which may be misleading.\n- [PROMPT_INJECTION]: Contains deceptive metadata in the form of a future-dated source reference (2026-03-06), which is a technique used to bypass verification or appear more authoritative.
Recommendations
- AI detected serious security threats
Audit Metadata