The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill uses a structured DataTable format to ingest data into VegaVisual and DataGrid components. If these components process data from untrusted external sources, there is a potential surface for indirect prompt injection, as the data influences the rendered output and interaction events.
Ingestion points: Data enters the visuals through the data prop in the VegaVisual and DataGrid components, as defined in SKILL.md and the references/data-table.md schema.
Boundary markers: The documentation does not specify the use of delimiters or specific instructions to the agent to ignore potentially malicious content embedded within the DataTable values.
Capability inventory: The components render interactive elements in the UI and emit events via the onInteraction prop (described in SKILL.md), which are then consumed by the host application.
Sanitization: There are no explicit instructions for the sanitization or validation of data values before they are passed to the visualization components.
[Integration with Trusted Libraries]: The skill utilizes official vendor packages including @microsoft/fabric-visuals, @microsoft/fabric-datagrid, and @microsoft/fabric-visuals-core. These are well-known libraries from a trusted organization used for their intended purpose of data presentation.
[Secure Link Implementation]: The example for rendering clickable URLs in references/data-grid-visual.md correctly implements the rel="noopener noreferrer" attribute, which is a standard security measure to prevent tab-napping and protect user privacy when navigating to external sites.

visuals