The Agent Skills Directory

[DATA_EXFILTRATION]: The skill supports an image element in scripts/svg_skill/core.py that accepts arbitrary href values from the input specification. When the resulting SVG is rendered using the cairosvg library in scripts/svg_skill/render.py, the renderer may attempt to access local files or internal network resources if the href points to sensitive paths or internal IPs.
[DYNAMIC_EXECUTION]: The validation component in scripts/svg_skill/validate.py uses the standard xml.etree.ElementTree library to parse SVG content. This parser is vulnerable to XML External Entity (XXE) attacks and XML-based denial-of-service (such as the 'Billion Laughs' attack) because it does not disable entity expansion by default.
[INDIRECT_PROMPT_INJECTION]: The skill processes structured JSON data that directly influences file system access and rendering engine behavior.
Ingestion points: Data enters via specification files read by scripts/svg_cli.py and processed in scripts/svg_skill/core.py.
Boundary markers: None; the skill lacks delimiters or instructions to prevent the processing of malicious content embedded within the specification.
Capability inventory: The skill provides capabilities for file writing, file reading, and rendering through cairosvg, which performs complex processing of XML and CSS.
Sanitization: While numeric values are validated, there is no sanitization of URI schemes or string content in href attributes, allowing the injection of dangerous file or network paths.

svg-precision-skill