motherduck-build-dashboard
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for running validation scripts using the
uvPython package manager (e.g.,uv run python skills/motherduck-build-dashboard/artifacts/dashboard_story_example.py). This is standard behavior for development and testing of agent skills within a local environment. - [DYNAMIC_EXECUTION]: The primary function of the skill is to generate and update "Dives," which are interactive React components rendered within the MotherDuck platform. This involves assembling TSX code blocks based on user requests and database schemas, which is then saved to the remote service.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data (database schemas and query results) to generate dashboard content. While this presents an indirect prompt injection surface where malicious data in the database could influence the generated dashboard's code or narrative, the risk is inherent to the skill's purpose as a data visualization tool. 1. Ingestion points: Data and metadata are retrieved from MotherDuck via
motherduck-exploreandmotherduck-query. 2. Boundary markers: None explicitly defined between data and prompt during SQL generation. 3. Capability inventory: The skill has the ability to write/update code (save_dive,update_dive) and execute SQL queries. 4. Sanitization: The skill relies on standard SQL and React best practices mentioned in the guides but does not implement custom sanitization for ingested data content.
Audit Metadata