create-sunpeak-app

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to clone the official Sunpeak repository from GitHub to access reference templates and working examples.
  • Evidence: git clone --depth 1 https://github.com/Alignbase/sunpeak /tmp/sunpeak targeting the author's infrastructure.
  • [EXTERNAL_DOWNLOADS]: The documentation provides a command to install an additional testing skill package from the author's official distribution channel.
  • Evidence: pnpm dlx skills add Alignbase/sunpeak@test-mcp-server used to provision environment-specific testing capabilities.
  • [INDIRECT_PROMPT_INJECTION]: The skill describes an architecture for building user interfaces that process data provided by an AI host, creating a surface for indirect prompt injection.
  • Ingestion points: Data from the AI host (such as tool inputs and outputs) enters the application context through specific framework hooks like useToolData in files such as src/resources/{name}/{name}.tsx.
  • Boundary markers: The framework uses React hooks and isolated component logic to manage data flow, and instructions recommend explicit Zod schema validation to enforce structure on ingested data.
  • Capability inventory: The application environment provides tools for calling server-side logic (useCallServerTool), sending messages to the chat context (useSendMessage), and initiating file downloads (useDownloadFile).
  • Sanitization: The instructions emphasize strict input and output validation using Zod schemas in tool definitions (src/tools/{name}.ts), which acts as a defense against malformed or malicious data payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 08:25 AM
Security Audit — agent-trust-hub — create-sunpeak-app