data-loading

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly requires the agent to "pass prefilled values the user already supplied" (which may include credentials) in the propose_connection action, forcing the LLM to include user-supplied secret values verbatim in its generated action payloads.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow ingests outsider-authored free text via the find_data/probe_data tool arguments (e.g., the user-provided regex query and other query fields) and passes them into DataDiscoveryService for execution without requiring prior selection of a specific item.