The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process User-Agent strings and HTTP headers from external, untrusted sources. This presents an indirect prompt injection surface if the parsed results (such as browser name, device type, or bot status) are subsequently used to construct prompts for an LLM without proper sanitization. The library itself provides some protection by trimming inputs to 500 characters to prevent ReDoS attacks.\n- [EXTERNAL_DOWNLOADS]: The instructions reference downloading the ua-parser-js library and its PRO variants from the official npm registry. It also mentions fetching a distribution build from the jsDelivr CDN and using a REST API provided in partnership with RESTCountries. These are established services and official distribution channels for the software described.\n- [COMMAND_EXECUTION]: The skill provides examples for using the library's CLI via bunx ua-parser-js. This is a standard method for executing utility scripts bundled with a package and is consistent with the skill's purpose as a developer tool.

ua-parser