The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides numerous templates and examples that ingest external data without explicit sanitization or boundary markers, creating a vulnerability surface.
Ingestion points: The data property in assets/chart-template.jsx, assets/interactive-template.jsx, and the data parameter in various drawing functions defined in SKILL.md and references/d3-patterns.md.
Boundary markers: Absent. The skill instructions do not advise the agent to use delimiters or treat input data as untrusted.
Capability inventory: The skill uses D3 to manipulate the DOM, including the .html() method for rendering tooltips in SKILL.md and assets/interactive-template.jsx. This method can render arbitrary HTML and execute scripts if the input data contains malicious payloads.
Sanitization: Absent. The provided code snippets directly interpolate data values (like d.label) into HTML strings for tooltips without escaping or validation.
[EXTERNAL_DOWNLOADS]: The skill documentation references the official D3.js library hosted on its well-known CDN for integration in vanilla JavaScript environments.
Evidence: Reference to https://d3js.org/d3.v7.min.js in SKILL.md.

d3-viz