connect-ai-manage-mcp
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill connects to
https://mcp.cloud.cdata.com/mcp/mgmt, which is the official Management MCP server for CData Connect AI. This is a legitimate vendor resource used for platform administration. - [COMMAND_EXECUTION]: Provides documentation and bash snippets for the user to run in their own local terminal. These commands assist in configuring the Claude CLI and securely generating local credentials (PATs) without exposing them to the AI agent.
- [CREDENTIALS_UNSAFE]: Implements a security guideline that explicitly forbids the agent from asking for or accepting passwords, API keys, or secrets in the chat. It uses browser-based OAuth for authentication, which is a security best practice.
- [INDIRECT_PROMPT_INJECTION]: The skill enables setting user-defined instructions for toolkits and individual tools through the
update_toolkit_server_instructionsandupdate_source_tool_instructionstools. This feature allows users to configure downstream agent behavior. - Ingestion points: User-defined instruction fields for toolkits and tools.
- Boundary markers: None explicitly defined within the skill; relies on the platform's tool execution environment.
- Capability inventory: Connection administration, toolkit configuration, and source operation management.
- Sanitization: Handled by the target platform's backend infrastructure.
Audit Metadata