rivet-dynamic-apps

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill's core functionality revolves around the dynamic generation and deployment of application code. The examples/apps-ai-builder/src/server.ts file uses an LLM to generate code for files like src/index.ts and package.json, which are then programmatically deployed using the deployApp function. Similarly, examples/apps-core-quickstart/src/server.ts demonstrates deploying a raw JavaScript string as a functional application.
  • [INDIRECT_PROMPT_INJECTION]: The builder example (examples/apps-ai-builder/src/server.ts) processes natural language prompts from users to generate application logic. The skill includes several mitigation strategies for this attack surface:
  • Ingestion Points: The /deploy/:name POST endpoint accepts a user-provided prompt.
  • Boundary Markers: The LLM is explicitly constrained via instructions to only edit specific files (package.json, tsconfig.json, src/index.ts).
  • Capability Inventory: The system can build and serve the generated code through the Rivet platform.
  • Sanitization: The implementation uses Zod for schema validation of the model's output and enforces strict limits on file sizes and the number of repair attempts.
  • [EXTERNAL_DOWNLOADS]: The documentation and examples reference several external resources and dependencies:
  • Fetches documentation and architectural guides from the vendor's official domain (rivet.dev).
  • References standard libraries such as hono, ai, and @ai-sdk/anthropic in package.json files.
  • Includes the vendor's own packages @rivet-dev/dynamic-apps and rivetkit.
  • [COMMAND_EXECUTION]: The skill documentation provides instructions for running and deploying the example applications using standard development tools like pnpm and npx tsx. The package.json scripts are standard for Node.js development environments.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 06:19 PM