artifacts-builder

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell scripts (init-artifact.sh and bundle-artifact.sh) to automate project setup and the build pipeline. This involves executing system commands such as pnpm, npm, tar, sed, and node to manipulate files and manage dependencies.
  • [EXTERNAL_DOWNLOADS]: The initialization and bundling scripts download a wide range of packages from the public NPM registry. These include well-known development tools and libraries like Vite, Parcel, Tailwind CSS, and Radix UI components. The downloads are standard for modern web development workflows.
  • [INDIRECT_PROMPT_INJECTION]: The skill environment involves processing untrusted inputs (user-supplied project names) and AI-generated source code through shell scripts and build processes.
  • Ingestion points: The project name provided to scripts/init-artifact.sh and the source code developed by the AI agent.
  • Boundary markers: No specific boundary markers are used to delimit generated content within the scripts.
  • Capability inventory: Shell execution, filesystem access for writing source code, and network access for package management.
  • Sanitization: Project name variables are quoted in shell scripts to prevent trivial command injection, and the build process uses standard compilers and bundlers.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 02:26 PM