design-system-builder
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands in Step 5 to verify the local development environment and suggest project initialization. This includes using
findto search forpackage.jsonfiles and searching for the string 'react' within them to detect existing projects. - [EXTERNAL_DOWNLOADS]: The environment setup instructions recommend the use of
npxandnpmto install packages includingtailwindcss,postcss,autoprefixer, andlucide-react. These downloads target the official npm registry, which is a well-known and trusted service for JavaScript development. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of user-provided reference images and project idea files.
- Ingestion points: Data enters the workflow in Step 1 through the reference images directory and the project idea file.
- Boundary markers: The skill utilizes structured markdown templates in the
assets/directory (e.g.,design-system.md,app-overview-generator.md) which use XML-style tags like<context>and placeholder tokens like{项目背景}to separate instructions from user-provided content. - Capability inventory: The skill has access to shell execution (
find,npm,npx) and the ability to write documentation and React component files to the local file system. - Sanitization: There is no explicit evidence of sanitization or character escaping for the ingested text files, relying on the agent's internal reasoning and boundary markers.
Audit Metadata