The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external platform (Twitter/X) and possesses several write and management capabilities, creating an attack surface for indirect prompt injection.
Ingestion points: Tools such as TWITTER_RECENT_SEARCH, TWITTER_FULL_ARCHIVE_SEARCH, and TWITTER_POST_LOOKUP_BY_POST_ID retrieve arbitrary text data from the public Twitter platform into the agent's context.
Boundary markers: The instructions lack specifications for using delimiters or explicit warnings to ignore embedded instructions within retrieved social media content.
Capability inventory: The skill provides high-privilege tools including TWITTER_CREATION_OF_A_POST, TWITTER_POST_DELETE_BY_POST_ID, TWITTER_ADD_POST_TO_BOOKMARKS, and TWITTER_UNLIKE_POST across its referenced scripts.
Sanitization: There is no documentation of sanitization, filtering, or escaping mechanisms for retrieved Twitter content before it is interpolated into agent prompts.
[EXTERNAL_DOWNLOADS]: The skill requires the configuration of an external MCP server at https://rube.app/mcp. This endpoint provides the underlying tool definitions and execution environment for the Twitter automation workflow.

twitter-automation