The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Python script (generate_chart_svg.py) for deterministic chart generation. The script implements robust security controls: it enforces absolute root paths, rejects symlink inputs or outputs, and prevents path traversal by verifying that all file operations occur strictly within the designated --root directory. It also uses atomic writes to prevent race conditions during file creation.
[DATA_EXPOSURE]: The skill processes research data into visual charts. It minimizes exposure risk by escaping all text content using html.escape and strictly validating URLs. Only http and https protocols are allowed for source links, preventing javascript: URI injections or other cross-site scripting (XSS) vectors within the generated SVG markup.
[INDIRECT_PROMPT_INJECTION]: While the skill processes untrusted research data (labels, values, and source URLs), it treats this data as non-executable content. The SKILL.md explicitly instructs agents to treat these fields as data and never as instructions. The generator script further enforces this by sanitizing all inputs and outputting only static SVG elements, with no support for <script> tags or event handlers.

blog-chart