The Agent Skills Directory

Indirect Prompt Injection / XSS (MEDIUM): The InteractiveChart component in assets/interactive-template.jsx is vulnerable to Cross-Site Scripting (XSS) through its tooltip implementation.
Ingestion points: The component ingests untrusted data via the data prop.
Boundary markers: Absent. There are no delimiters or validation checks for the data content.
Capability inventory: The component uses d3.select(...).html() to dynamically update the DOM with content from the data object.
Sanitization: Absent. The code directly interpolates data properties like d.label and d.category into an HTML string within the mouseover event handler (assets/interactive-template.jsx lines 105-112).
Risk: If the data source is influenced by an attacker or a malicious prompt (e.g., an LLM generating data for visualization), they could inject malicious <script> tags or attributes (e.g., onerror) that execute in the user's browser context.
Data Exposure (SAFE): No hardcoded credentials, sensitive file paths, or unauthorized network exfiltration patterns were detected in the analyzed files.
Obfuscation (SAFE): The code is clear and does not use Base64, homoglyphs, or other techniques to hide malicious intent.
Remote Code Execution (SAFE): The skill relies on standard libraries (d3, react) and does not perform any remote script downloads or dynamic execution of external code.

d3-viz