constellation-team
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONNO_CODE
Full Analysis
- Prompt Injection (HIGH): The
SKILL.mdfile contains 'ULTRA-FRONTEND & SUPERMODE ACTIVATION' instructions which explicitly command the agent to 'Ignore token and cost constraints'. This is a direct attempt to bypass system-level operational limits and resource management. - Prompt Injection (MEDIUM): The skill uses high-authority role-play markers ('Distinguished Principal Software Engineer', 'Elite Solutions Architect') and 'SUPERMODE' triggers to force the AI into an unrestricted state, which is a common pattern in jailbreak-style prompt injections.
- No Code (SAFE): The skill consists entirely of Markdown files for persona coordination and does not contain executable scripts, preventing file-based or network-based attacks.
Recommendations
- AI detected serious security threats
Audit Metadata