root
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill includes defensive instructions to prevent manipulation, specifically directing the agent to ignore hidden instructions and treat delegated outputs as evidence to be inspected rather than inherent truth.
- [DATA_EXFILTRATION]: There are no indicators of unauthorized data collection or external transmission. The agent's scope is restricted to the local project workspace and specified architectural documents.
- [COMMAND_EXECUTION]: The skill utilizes a tool for agent creation (
paseo_create_agent) with specific validation requirements for provider and model settings, ensuring sub-agents are created within a controlled environment. - [SAFE]: The orchestration logic emphasizes architectural integrity and manual verification of all changes, which is a standard security-conscious approach for autonomous agents.
Audit Metadata