NYC

constraints-scheduling

Fail

Audited by Gen Agent Trust Hub on Feb 15, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • Indirect Prompt Injection (HIGH): The skill requires the agent to ingest untrusted external data (calendar/ICS files) and then perform execution tasks (writing and running verification scripts). This combination allows for a scenario where malicious content in a calendar file could influence the agent's logic or the scripts it generates. Ingestion points: External ICS/calendar files described in Phase 1. Boundary markers: No specific delimiters or instruction-guards are defined for the data processing phase. Capability inventory: Phase 3 and Best Practices mandate writing and executing scripts (subprocess/exec capability). Sanitization: Phase 1 includes an integrity check and recommends proper parsers, which mitigates file-level corruption but does not prevent content-based prompt injection.
  • Dynamic Execution (LOW): The skill instructs the agent to generate and run scripts for constraint checking. While appropriate for the task, automated code generation and execution is a form of dynamic execution that increases the overall risk profile when combined with external inputs.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 15, 2026, 11:30 PM