Skills
skills/qu-skills/skills/building-inferencesh-apps/Gen Agent Trust Hub

building-inferencesh-apps

Fail

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides installation commands for the platform's CLI using a curl | sh pattern from cli.inference.sh.- [REMOTE_CODE_EXECUTION]: Recommends installing development prerequisites such as 'uv', 'fnm', and 'nvm' using shell-piped installation scripts from their respective official domains and GitHub repositories.- [EXTERNAL_DOWNLOADS]: Refers to downloading AI models and software dependencies from Hugging Face, npm, and PyPI.- [PROMPT_INJECTION]: The skill documents the creation of applications that ingest user prompts, which creates an indirect prompt injection surface.
  • Ingestion points: Untrusted prompt data enters the application through AppInput (Python) or RunInput (Node.js) models as seen in SKILL.md and logic reference files.
  • Boundary markers: Code templates do not demonstrate the use of delimiters or 'ignore' instructions for the prompt data.
  • Capability inventory: The documented applications have capabilities to perform network requests (httpx, OpenAI), access the file system (File class), and execute model-related logic.
  • Sanitization: The templates rely on platform-level validation and model-side guardrails rather than explicit content sanitization.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 12, 2026, 10:27 PM
Security Audit — agent-trust-hub — building-inferencesh-apps