research-engineer
Pass
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: LOW
Full Analysis
- Prompt Injection (LOW): The skill utilizes strong persona-shaping language ("You are not an assistant", "Zero-Hallucination Mandate"). While these are directive, they are used to establish professional boundaries and scientific rigor rather than bypassing safety protocols or instructions to reveal system prompts.
- Indirect Prompt Injection (LOW): The skill is designed to ingest and critique user-provided code and theoretical premises. This represents a standard attack surface for an AI agent. However, the instructions focus on generating critiques and implementations rather than executing side effects, and the skill lacks any commands for network exfiltration or unauthorized file system access.
- Unverifiable Dependencies (INFO): The skill references various libraries (e.g., lxml, gumbo, std::atomic) and languages (Rust, Ada, Fortran) as recommendations for the AI to provide in its output. It does not provide automated scripts to download or execute these dependencies from untrusted sources.
- General Security Posture (SAFE): Analysis of the instructions found no evidence of obfuscated code, hardcoded credentials, persistence mechanisms, or attempts to escalate privileges. The implementation standards provided (e.g., atomic operations, memory safety) align with secure development practices.
Audit Metadata