skills/eszxcvfd/skills/root/Gen Agent Trust Hub

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
Risk Level
SAFE
Analyzed
Aug 10, 2026, 03:24 AM
Security Audit — agent-trust-hub — root