The Agent Skills Directory

[SAFE]: The skill generates static HTML visualizations and uses a local script scripts/validate.py to ensure the output conforms to safety and design constraints.
[REMOTE_CODE_EXECUTION]: The validation script is executed via uv run, which is a standard and safe way to run local Python tools within a managed environment. No remote scripts are executed.
[EXTERNAL_DOWNLOADS]: The skill guidelines and the validation script enforce a strict allowlist of trusted CDN hosts (e.g., cdnjs, jsdelivr, cloudflare) for loading libraries like Tailwind CSS and Chart.js. This is documented and restricted to prevent loading resources from arbitrary or untrusted domains.
[COMMAND_EXECUTION]: The execution of the validation script uv run scripts/validate.py <output-path> is restricted to the specific local file generated by the agent, posing no risk of command injection or arbitrary execution.
[DATA_EXFILTRATION]: The skill guidelines explicitly forbid non-CDN external requests, and the validation script enforces this by checking all resource URLs against a safe list. No sensitive data access or exfiltration patterns were detected.

visualise