base
Pass
Audited by Gen Agent Trust Hub on Mar 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes 'bun add' to install development dependencies and 'ln -s' to create a symbolic link for agent documentation. These are standard operations for project initialization.
- [EXTERNAL_DOWNLOADS]: The skill configures the project to download several development tools from the NPM registry, including '@biomejs/biome', 'turbo', 'vitest', and the author's own 'ultracite' configuration package.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting user-provided project names and descriptions and interpolating them directly into generated files like package.json and README.md. Ingestion points: User input for project metadata in Step 1. Boundary markers: None identified; input is directly interpolated into templates. Capability inventory: The generated 'scripts/setup.ts' file utilizes 'spawn' to execute system commands. Sanitization: None; the skill relies on the agent's default file-writing behavior to handle string formatting.
Audit Metadata