testing-skills-with-subagents
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- PROMPT_INJECTION (HIGH): The documentation contains emphatic and directive language aimed at overriding the agent's default reasoning and safety protocols.
- Evidence: Use of markers like
IMPORTANT: This is a real scenario. Choose and act.andTHIS IS EXTREMELY IMPORTANT. BEFORE ANY TASK, CHECK FOR SKILLS! - Evidence: Explicit attempts to bypass agent reasoning:
If a skill existed for your task and you didn't use it, you failed.andAgent can't rationalize away compliance. - Evidence: The file structure mimics known 'jailbreak' or behavioral override techniques by creating artificial urgency ('Production is bleeding money') to prioritize file-based instructions over standard operational logic.
Recommendations
- AI detected serious security threats
Audit Metadata