building-inferencesh-apps
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill provides instructions to download and execute installers for platform-specific and general development tools. This includes the
beltCLI (cli.inference.sh) and tools such asuv(astral.sh),fnm(fnm.vercel.app), andnvm(raw.githubusercontent.com/nvm-sh). These are standard installation procedures for the targeted development environment and originate from the platform vendor or well-known services. - [COMMAND_EXECUTION]: The skill uses the
beltCLI tool for managing the application lifecycle, including project initialization (belt app init), testing (belt app test), and deployment (belt app deploy). These operations are consistent with the skill's primary purpose of application development. - [INDIRECT_PROMPT_INJECTION]: The skill documents patterns for developing apps that ingest untrusted user data through
AppInputschemas. While it lacks explicit natural language boundary markers in prompt templates, it mandates structural sanitization using Pydantic and Zod validation. The capability inventory includes file system access and network operations, making proper validation critical for security.
Audit Metadata