The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the web but implements protective measures. Ingestion points: External search results and webpage content are fetched by scripts/search.js and scripts/content.js. Boundary markers: The scripts use clear delimiters like --- Result --- and --- Content --- in the console output to distinguish untrusted content for the agent. Capability inventory: The skill is limited to performing network requests and console output; it has no file system write or arbitrary command execution capabilities. Sanitization: The skill uses the turndown library to convert HTML to Markdown, which strips potentially malicious tags such as <script>, <style>, and other active elements before the content is presented to the agent.\n- [EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch search results from a configured SearXNG instance and page content from external URLs. This behavior is consistent with the skill's primary purpose and does not involve downloading or executing external scripts.\n- [COMMAND_EXECUTION]: Network operations are handled via standard fetch APIs. No dangerous patterns such as eval(), exec(), or dynamic shell command construction were found.

searxng-search