skills/carapace-sh/carapace-bin/brew/Gen Agent Trust Hub

brew

Pass

Audited by Gen Agent Trust Hub on Sep 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for executing arbitrary shell commands and Ruby code through Homebrew's native interfaces. Key examples include:
  • The brew ruby -e and brew exec commands, which allow direct execution of Ruby and shell code within the Homebrew environment (referenced in references/developer.md).
  • The postinstall stanza in Brewfiles and Formulae, which executes shell strings after package installation (referenced in references/bundle.md and references/formula.md).
  • The system call within the Formula DSL, used to execute build steps like make install (referenced in references/formula.md).
  • [PRIVILEGE_ESCALATION]: The skill documents operations requiring elevated permissions, which is standard for system-level package management:
  • Use of sudo to manage system-wide services via brew services (referenced in references/services.md).
  • Installation of .pkg files via the macOS installer, which typically requires administrative privileges (referenced in references/cask.md).
  • Use of chmod and ownership modification commands within cask flight blocks (referenced in references/cask.md).
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a significant attack surface for indirect prompt injection as it processes executable configuration files from potentially untrusted sources:
  • Ingestion points: The skill is designed to process user-provided Brewfile, Formula (.rb), and Cask (.rb) files, as well as metadata from remote URLs during installation (referenced in SKILL.md and references/bundle.md).
  • Boundary markers: The provided documentation does not include explicit warnings or delimiters to differentiate between trusted and untrusted configuration files.
  • Capability inventory: The agent has the capability to execute shell commands (system, postinstall), run Ruby code (brew ruby), and perform network operations (brew install, brew tap).
  • Sanitization: There is no mention of sanitizing or validating the content of user-provided Ruby DSL files before evaluation.
  • [EXTERNAL_DOWNLOADS]: The skill documents how Homebrew fetches binary bottles, source code, and API metadata from external services:
  • Downloads bottles from GitHub Container Registry (ghcr.io/v2/homebrew/core), a well-known and trusted service (referenced in references/bottle.md).
  • Fetches formula and cask metadata from the official Homebrew API (formulae.brew.sh/api) (referenced in references/taps.md).
  • [DYNAMIC_EXECUTION]: The skill documents Homebrew's core design of evaluating Ruby-based Domain Specific Languages (DSLs) at runtime to define installation logic (referenced in references/formula.md and references/cask.md). This includes the use of Ruby logic (conditionals and variables) within Brewfile configurations.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 30, 2026, 08:51 PM