ai-infrastructure-modal

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The reference.md file demonstrates a pattern for image construction using .run_commands("curl https://example.com | bash"). Piped remote execution is a high-risk practice that can lead to supply chain attacks if the source is compromised.
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of GPU-backed API endpoints that process untrusted user data.
  • Ingestion points: External data enters through function parameters in inference.py and Classifier.classify within examples/core.md.
  • Boundary markers: None present. The instructions do not recommend the use of delimiters or system prompts to isolate user-provided text from model instructions.
  • Capability inventory: The environment supports shell command execution and subprocess spawning (subprocess.Popen).
  • Sanitization: The provided examples lack input validation or filtering to prevent malicious payloads from influencing model behavior or exploiting the underlying serverless environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 03:57 PM
Security Audit — agent-trust-hub — ai-infrastructure-modal