Skills
skills/ldmrepo/michael/a2a-protocol/Gen Agent Trust Hub

a2a-protocol

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • Remote Code Execution (CRITICAL): The skill executes raw user input using the Python eval() function.
  • Evidence: In examples/simple_calculator_agent.py, the execute method calls eval(user_input).
  • Impact: Attackers can execute arbitrary Python commands, allowing for total control over the environment and the underlying host system.
  • Indirect Prompt Injection (LOW): The skill lacks input sanitization for user-provided data processed by a high-risk capability.
  • Ingestion point: context.get_user_input() in examples/simple_calculator_agent.py.
  • Boundary markers: Absent.
  • Capability inventory: eval() call in examples/simple_calculator_agent.py.
  • Sanitization: Absent.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 17, 2026, 06:23 PM