The Agent Skills Directory

[Data Management and Connector Handling]: The skill is designed to browse connected data sources and add new connectors via user-confirmed forms. It handles 'prefilled' values which may include user-supplied credentials. The implementation includes logic to process these values as transient form seeds, which is a standard approach for managing connection configuration.
[Credential Safety Instructions]: In its instructions, the skill explicitly directs the agent not to repeat user-supplied credentials in prose or subsequent tool output. This instruction is a security-conscious measure to prevent the accidental exfiltration or exposure of sensitive data within the conversation history.
[External Data Ingestion Surface]: The skill interacts with external data sources through tools like probe_data and describe_data. This introduces a potential surface for indirect prompt injection if the external source contains malicious instructions. However, the skill mitigates this by applying grounding rules that require all operations to be based on discovered metadata and by keeping queries structured and bounded, rather than generating raw SQL.

data-loading