The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided data files (CSV, JSON, Excel), creating a surface for indirect prompt injection where malicious content within those files could attempt to influence the agent's behavior during the analysis phase.
Ingestion points: Untrusted data enters the agent's context through scripts/analyze_data.py and scripts/prepare_data.py, which read files from the /mnt/user-data/uploads/ directory.
Boundary markers: The instructions do not specify the use of clear delimiters or 'ignore embedded instructions' markers when the agent reads or analyzes the content of the data files.
Capability inventory: The skill utilizes python for data analysis and bash (specifically via heredoc) for writing JSX files to the file system.
Sanitization: Data is sanitized using json.dumps() before being embedded into the final React artifact as an inline data island.
[EXTERNAL_DOWNLOADS]: The assets/components/ChartExplorer.jsx template specifies downloading the Vega, Vega-Lite, and Vega-Embed libraries from the official jsDelivr CDN (https://cdn.jsdelivr.net/npm/). These are well-known, industry-standard libraries used for data visualization.
[DYNAMIC_EXECUTION]: The skill generates JSX files by injecting data and chart specifications into a template at runtime. Additionally, the generated React artifact perform dynamic script loading to include necessary visualization libraries.
[COMMAND_EXECUTION]: The skill executes local Python scripts and shell commands to analyze data and manage output files. This involves passing user-supplied file paths as arguments to command-line tools.

charting-vega-lite