building-web-artifacts

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/init-artifact.sh script is vulnerable to command injection through the project name argument. The variable is concatenated into a sed command string without proper sanitization, allowing for shell breakout if the name contains specific characters like single quotes or semicolons.
  • [EXTERNAL_DOWNLOADS]: The skill performs multiple package installations using pnpm and npm, including a global installation of the pnpm tool. It downloads a wide range of standard frontend libraries (React, Vite, Tailwind CSS, Radix UI) from the official registries.
  • [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface where an agent, instructed by a user to use a malicious project name, could inadvertently execute arbitrary commands on the host system during the project initialization phase.
  • Ingestion points: The project name argument in scripts/init-artifact.sh.
  • Boundary markers: None present; the argument is used directly in shell command construction.
  • Capability inventory: The skill has the ability to execute shell commands, install packages, and write to the filesystem.
  • Sanitization: Missing sanitization for the project name argument before shell interpolation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 11:03 AM
Security Audit — agent-trust-hub — building-web-artifacts