The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard utilities (curl, jq, wc) and platform-specific tools (secretcurl, notify) for its operations. Payload construction for API calls is handled safely using jq to avoid shell injection.
[EXTERNAL_DOWNLOADS]: Data is fetched from the official X.AI API and through general web search/fetch tools to gather information on the target subject. These are legitimate operations for the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from X and the web (ingestion points). It addresses this risk with explicit boundary instructions: the agent is told to treat content as data only and to discard any embedded directives like 'ignore your instructions' (sanitization/boundary markers). Its capabilities include network access and file system writes, but these are scoped to the intended profile-building task.

soul-builder