building-inferencesh-apps

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). The list includes multiple remote install scripts (curl|sh or PowerShell installers) hosted on project-specific domains (cli.inference.sh, astral.sh, fnm.vercel.app) which are direct executable downloads that can be abused to distribute malware if the source or supply chain is compromised, so they should be treated as suspicious until verified.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). The runtime workflow documented here only consumes outsider-authored free text when it is provided explicitly as app request input (e.g., belt app run ... --input '{"prompt": "hello"}'), not from any monitored queue/feed, store browsing, or automatic ingestion of external free text.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill includes installation commands that curl/pipe remote shell scripts and execute them (e.g., https://cli.inference.sh, https://astral.sh/uv/install.sh, https://fnm.vercel.app/install, https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.0/install.sh), which fetch and run remote code as required setup steps for the CLI/required tooling.