neo-agentic-design
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of Markdown documentation and JSON evaluation files. It provides architectural patterns for LLM orchestration without providing any functional code or implementation scripts.
- [PROMPT_INJECTION]: The skill does not attempt to override agent behavior. On the contrary, Chapter 18 ('Guardrails & Safety Patterns') in
references/advanced-safety.mdprovides explicit instructions on how to detect and prevent prompt injection and jailbreak attacks. - [DATA_EXFILTRATION]: No exfiltration patterns detected. The designs for memory management and inter-agent communication (A2A) are described abstractly using standard protocols like JSON-RPC and HTTP.
- [NO_CODE]: There are no executable files (e.g., .py, .js, .sh) provided in the skill package. All logic is conceptual and framework-agnostic.
- [COMMAND_EXECUTION]: There is no use of the dynamic context injection syntax (
!command) or any instructions for the agent to execute shell commands. Systems that call tools or execute code are described as sandboxed environments in the reference documents.
Audit Metadata