stitch-react-components
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses a bash script (
scripts/fetch-stitch.sh) to download HTML design files from Google Cloud Storage URLs provided by the Stitch MCP tool. This operation is used for high-reliability fetching of design assets. - [COMMAND_EXECUTION]: The agent is instructed to execute local shell commands and Node.js package manager commands (
npm install,npm run dev,npm run validate) to set up the development environment and perform quality checks on generated components. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) because it processes untrusted external design data (JSON/HTML) retrieved from the Stitch API to generate React code.
- Ingestion points: Design data is retrieved via
stitch-mcp-get-screenand thefetch-stitch.shscript. - Boundary markers: None identified in the prompt templates.
- Capability inventory: Includes file writing, shell script execution, and Node.js command execution.
- Sanitization: Relies on the agent's internal logic and an optional validation script (
npm run validate).
Audit Metadata