find-key-internal-sources
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- Internal Data Integration: The skill interacts with internal systems such as CRM records, document stores, and messaging history to identify relevant information. This access is central to the skill's functionality, and it includes specific rules to respect privacy, such as defaulting to public channels and excluding inactive accounts.
- User-Directed Actions: While the skill can draft communications (e.g., Slack messages or handoff notes), it is explicitly restricted from posting or sharing them without direct user approval. This ensures that the user remains in control of all external interactions.
- Information Processing Surface: The skill processes various forms of input, including meeting notes and internal message threads. While this creates a standard surface for indirect prompt injection common to text-processing AI, the skill's architecture includes request normalization and boundary-setting steps to maintain focus on its routing task.
- Scoped Configuration: The skill utilizes a preflight process to load configuration and context from a related internal skill. This modular approach helps maintain consistent security policies and user context across different tasks.
Audit Metadata