The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from Wikipedia's public APIs (including page summaries, extracts, and search snippets) which are external and untrusted sources. This creates a surface where malicious content on Wikipedia could theoretically influence the agent's behavior.
Ingestion points: Wikipedia API responses as described in references/topic-search.md and references/article-next-steps.md.
Boundary markers: The instructions do not define specific delimiters or explicit warnings to the agent to disregard instructions found within the fetched Wikipedia content.
Capability inventory: The skill has the ability to trigger navigation events in the user's environment via the message_parent tool.
Sanitization: The skill contains a strong domain restriction, explicitly requiring that all url_change targets remain within the *.wikipedia.org domain.
[EXTERNAL_DOWNLOADS]: The skill interacts with well-known Wikipedia API endpoints (en.wikipedia.org) to fetch article metadata and search results. These references to established services are considered safe within the context of the skill's primary function.

wikipedia-discovery