claude-d3js-skill
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides code templates that render HTML based on input data without sanitization, creating a Cross-Site Scripting (XSS) vulnerability surface.
- Ingestion points: The
dataparameter indrawVisualization(SKILL.md) and theBasicChart/InteractiveChartcomponents (assets/chart-template.jsx,assets/interactive-template.jsx). - Boundary markers: Absent; data is interpolated directly into template literals.
- Capability inventory: The code uses the D3
.html()method to inject strings into the DOM, which can execute scripts if the data is malicious. - Sanitization: No sanitization or escaping logic is present for data labels or categories before DOM injection.
- [REMOTE_CODE_EXECUTION]: Fetches the D3.js library from the official d3js.org content delivery network (CDN).
- Evidence:
<script src="https://d3js.org/d3.v7.min.js"></script>inSKILL.md. - [METADATA_POISONING]: The skill frontmatter contains a
risk: criticalfield which is inconsistent with the functional nature and primary purpose of the skill. - Evidence:
risk: criticallabel inSKILL.mdYAML frontmatter.
Audit Metadata