NYC

artifacts-builder

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The skill requires the execution of scripts/init-artifact.sh and scripts/bundle-artifact.sh. These scripts are not provided in the analyzed content, preventing a full audit of their local system operations.
  • EXTERNAL_DOWNLOADS (MEDIUM): The documentation for scripts/bundle-artifact.sh explicitly states it installs packages like parcel, @parcel/config-default, and html-inline from the public npm registry at runtime.
  • REMOTE_CODE_EXECUTION (MEDIUM): The skill involves running a build toolchain (Parcel/Vite) on code that is dynamically generated or modified based on user input, which could potentially lead to code execution if malicious patterns are injected into the source files.
  • PROMPT_INJECTION (LOW): Indirect prompt injection surface exists through the processing of user-defined artifact requirements.
  • Ingestion points: User instructions for artifact UI/logic.
  • Boundary markers: Absent.
  • Capability inventory: Shell command execution (bash) and network access (npm).
  • Sanitization: None identified in the provided skill text.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 05:41 PM