skills/carlheath/ogmios/engineering/Gen Agent Trust Hub

engineering

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • PROMPT_INJECTION (HIGH): The skill contains explicit instructions to bypass standard user consent protocols, stating "NO user confirmation needed for TIER 1 skills. Activate immediately." This is a direct attempt to override the agent's safety and orchestration layer.
  • COMMAND_EXECUTION (HIGH): The skill possesses high-privilege capabilities for code implementation and bug fixing. It is designed to ingest untrusted data (user code and error logs) without defined boundary markers or sanitization, creating a high-risk surface for indirect prompt injection where an attacker could manipulate the agent into writing or executing malicious code.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 10:15 AM