html-data-explorer
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided data (CSVs, JSON, logs) to generate an interactive HTML explorer. This creates a vulnerability surface for indirect injection if the data contains malicious content.
- Ingestion points: Dataset content (tabular or log data) is ingested from the agent's context into the generated HTML file.
- Boundary markers: The skill does not explicitly define boundary markers for the data content but relies on secure implementation patterns.
- Capability inventory: The generated artifact is self-contained and operates entirely in the browser (no-fetch rule). The agent generates the file using platform-provided tools.
- Sanitization: The skill enforces strong mitigations, including a mandatory secret-redaction pass using comprehensive regex patterns for API keys and tokens. It also mandates safe DOM construction using
textContentandcreateElementto prevent Cross-Site Scripting (XSS). - [EXTERNAL_DOWNLOADS]: The skill allows the use of specific, well-known data visualization libraries (Chart.js and D3) from pinned CDN links as a scoped exception to the general policy.
- These downloads are restricted to reputable libraries intended only for the primary purpose of data charting.
Audit Metadata