The Agent Skills Directory

[Indirect Prompt Injection] (SAFE): The skill ingests external data from CSV files and outputs it into HTML reports.
Ingestion points: scripts/main.py reads user-provided CSV files via csv.DictReader.
Boundary markers: None used in the prompt or data processing.
Capability inventory: Local file writing via Path.write_text.
Sanitization: Data values are converted to floats and chart data is serialized via json.dumps. However, column headers (labels) are interpolated into the HTML without explicit escaping, creating a potential XSS surface if a malicious CSV is processed and the resulting HTML is opened in a browser.
[External Downloads] (SAFE): The generated HTML includes a script tag for chart.js from cdn.jsdelivr.net. This is a standard and widely trusted CDN for frontend libraries.
[Metadata Poisoning] (SAFE): The SKILL.md file lists plotly as a dependency, but it is not used in the implementation. This appears to be a minor documentation oversight rather than a deceptive practice.

data-visualizer