Skills
Navigate
skills/inference-sh/skills/building-inferencesh-apps/Gen Agent Trust Hub

building-inferencesh-apps

Fail

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions to download and execute the vendor's belt CLI tool using curl -fsSL https://cli.inference.sh | sh in SKILL.md and references/cli.md.
  • [REMOTE_CODE_EXECUTION]: Includes instructions to install development tools from well-known services via shell-piped execution, such as uv from astral.sh, fnm from fnm.vercel.app, and nvm from raw.githubusercontent.com.
  • [INDIRECT_PROMPT_INJECTION]: The application framework described in the skill is designed to ingest and process external data, creating an attack surface.
  • Ingestion points: Data enters via the run() method through AppInput (Python) and RunInput (Node.js) models defined in SKILL.md and references/node-app-logic.md.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are included in the provided code templates.
  • Capability inventory: Apps built using this logic can perform network operations (e.g., via httpx), file system writes (to /tmp), and execute CLI commands via the belt tool.
  • Sanitization: The provided boilerplate logic does not include sanitization or escaping of input data before processing.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 15, 2026, 12:42 PM