install-dependency

Fail

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: HIGHPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The scripts/install-js.sh script downloads and executes the Bun installation script from https://bun.sh/install using a shell pipe (curl | bash). This is a common pattern for installing this specific runtime, which originates from a well-known service.
  • [PRIVILEGE_ESCALATION]: The scripts/install-system.sh script employs sudo to install packages via system package managers such as apt, brew, and dnf. This allows the agent to perform administrative operations on the host system. The skill includes a manual check for an APPROVED=1 environment variable and documentation instructing the agent to seek user approval before proceeding.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute a suite of shell scripts for environment setup, package scanning, installation, and verification. These scripts interact with the filesystem, environment variables, and external package managers.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes package names provided through agent instructions or project data, which are then passed to package managers with high privileges. This creates a surface for indirect prompt injection if an attacker can influence the package names the agent attempts to install.
  • Ingestion points: Package names are ingested in SKILL.md and passed as arguments to scripts like scripts/install-python.sh and scripts/install-js.sh.
  • Boundary markers: No specific delimiters or boundary markers are used when interpolating package names into shell commands.
  • Capability inventory: The skill possesses the ability to execute commands with sudo, perform network downloads, and modify the filesystem via package managers.
  • Sanitization: While package names are double-quoted in shell scripts to prevent basic word splitting, there is no validation or sanitization to prevent the installation of malicious packages or command-line flag injection.
  • [EXTERNAL_DOWNLOADS]: The skill performs external network requests to download the Bun installer and fetch packages from public registries like PyPI and npm during its normal operation.
Recommendations
  • HIGH: Downloads and executes remote code from: https://bun.sh/install - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 30, 2026, 05:10 PM