carto-render-inline-map
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves data from external database connections via the explore_data tool to compose declarative map specifications. This creates an indirect prompt injection surface where malicious data in the source tables could potentially influence the agent's behavior or output.
- Ingestion points: Data enters the agent's context through the explore_data tool (specifically the describe and list_resources methods) to obtain table schemas and column statistics as described in both SKILL.md and references/cartography.md.
- Boundary markers: No specific delimiters or safety instructions are defined to separate the fetched database content from the agent's instructions during the composition of the map spec.
- Capability inventory: The skill utilizes the view_map tool to render interactive maps and the show_widget tool to display legends.
- Sanitization: The instructions note that chat-message HTML is escaped by host renderers, which serves as a mitigation against script injection in the user interface.
Audit Metadata