The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exposes a potential attack surface by processing untrusted data and rendering it directly into the DOM without sanitization.
Ingestion points: External data is ingested through the data parameter in various visualization functions in SKILL.md and through the data prop in the React components provided in assets/chart-template.jsx and assets/interactive-template.jsx.
Boundary markers: The skill does not implement boundary markers or instructions to the agent to ignore potentially malicious content embedded in the data streams.
Capability inventory: The skill utilizes the D3 .html() method to render tooltips (e.g., in SKILL.md line 489 and assets/interactive-template.jsx line 118). This method provides the capability to render and execute HTML/scripts if the input data is not trusted.
Sanitization: There is no evidence of data sanitization, escaping, or schema validation performed on the ingested data before it is interpolated into the DOM via the .html() method.
[EXTERNAL_DOWNLOADS]: The skill references an external resource for its functionality.
It includes instructions to load the d3.js library from its official CDN (https://d3js.org/d3.v7.min.js). This is a reference to a well-known and established open-source library.

d3-viz