connect-ai-onboarding-claude-code
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use platform-specific commands ('open', 'Start-Process', 'xdg-open') to help users access official CData sign-up and connection management pages in their browser. It also uses the Claude CLI ('claude mcp add') to register the Connect AI MCP server.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from the 'getCatalogs' tool and user inputs during the onboarding sequence. * Ingestion points: User input for state selection in Step 0 and the output of the 'getCatalogs' tool in Step 4. * Boundary markers: Not present. * Capability inventory: Local command execution for browser interactions and CLI configuration. * Sanitization: Not present.
- [SAFE]: The skill provides defensive instructions for handling authentication, specifically guiding the user to process tokens in their local shell to ensure that no long-lived secrets are exposed in the AI conversation transcript.
Audit Metadata