skills/barefootford/buttercut/setup/Gen Agent Trust Hub

setup

Warn

Audited by Gen Agent Trust Hub on Oct 4, 2026

Risk Level: MEDIUMPERSISTENCEPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PERSISTENCE]: The skill performs persistent modifications to several shell configuration files, including ~/.zshrc, ~/.zshenv, ~/.zprofile, ~/.bash_profile, and ~/.bashrc. It appends environment variable exports and activation hooks for the mise version manager to ensure the development environment remains active across all shell session types.
  • [PRIVILEGE_ESCALATION]: Includes instructions for the agent or user to execute commands with sudo to remove system-level directories in /Library/Developer/CommandLineTools during troubleshooting.
  • [REMOTE_CODE_EXECUTION]: Facilitates the execution of the Homebrew installation script by downloading it from GitHub and piping the output directly into a bash shell.
  • [EXTERNAL_DOWNLOADS]: Fetches software components from external sources, including the Homebrew install script, Python packages (whisperx, pyannote-audio) via pip, and Ruby/Python binaries via the mise version manager.
  • [DYNAMIC_EXECUTION]: Generates a shell wrapper script at ~/.buttercut/whisperx at runtime and applies executable permissions using chmod +x.
  • [COMMAND_EXECUTION]: Executes a series of system commands for environment verification, package installation, library linking, and tap configuration for FFmpeg.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: Processes user inputs for installation type preference and error reporting consent via AskUserQuestion and external script output.
  • Boundary markers: Absent; the skill does not use specific delimiters to isolate external inputs or user preferences from instructions.
  • Capability inventory: Includes capabilities to write to sensitive shell profiles, execute arbitrary shell commands, and request administrative access via sudo.
  • Sanitization: No explicit sanitization or validation is applied to user-provided data or tool outputs before they are used in shell commands or configuration files.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 4, 2026, 03:05 PM