The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the xurl command-line utility to query account information and retrieve recent posts for style analysis. It specifically invokes xurl whoami and xurl posts or direct API endpoint queries to retrieve data from X (Twitter).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external source (X/Twitter posts) to perform style matching, which introduces a vulnerability to indirect prompt injection.
Ingestion points: User's recent posts are fetched via xurl posts @handle or direct API calls to the Twitter v2 API endpoints within SKILL.md.
Boundary markers: There are no explicit delimiters or boundary markers used to separate the fetched content from the agent's instructions.
Capability inventory: The skill uses xurl for read-only operations by default but has the capability to perform write operations (posts, replies) if the user explicitly requests them.
Sanitization: The skill is instructed to focus on linguistic patterns rather than subject matter, which provides a conceptual layer of separation, but it lacks technical sanitization for the content being processed.

draft-tweet