sdmx-explorer
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill guides the installation of the
opensdmxCLI tool viauv tool installand DuckDB through its official website. Theopensdmxtool is a resource developed by the skill's author, and DuckDB is a widely recognized service. - [REMOTE_CODE_EXECUTION]: Documentation in the skill suggests installing DuckDB by piping a script from its official domain (
https://install.duckdb.org) to the shell. While this is a sensitive execution pattern, it targets a well-known and established service. - [INDIRECT_PROMPT_INJECTION]: The skill interacts with and processes data from multiple external SDMX providers, creating a potential surface for indirect prompt injection.
- Ingestion points: Statistical data is fetched from providers like Eurostat, ISTAT, and the World Bank as described in
SKILL.mdandistat-flow.md. - Boundary markers: There are no explicit instructions for using delimiters or warnings to ignore instructions that might be embedded in the retrieved statistical datasets.
- Capability inventory: The skill utilizes the
opensdmxCLI for data operations, writes files (CSV, YAML, README) to the local filesystem, and recommends generating and running Python scripts for visualization. - Sanitization: The instructions do not specify any sanitization or validation logic for the external data being processed.
- [DYNAMIC_EXECUTION]: The skill encourages the agent to generate and execute Python scripts using the
plotninelibrary to create custom visualizations that go beyond the capabilities of the CLI tool.
Audit Metadata