The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from Wikipedia, a publicly editable platform, through tools like get_content, get_summary, and search_pages (identified in SKILL.md). This data is untrusted and could contain hidden instructions intended to manipulate the agent. The instructions do not include boundary markers or warnings to treat the fetched content as untrusted. Furthermore, the skill documentation suggests integration with powerful tools like pyats-routing and pyats-security which possess network device access capabilities, increasing the potential impact of a successful injection.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via python3 $MCP_CALL to perform its functions. This is the primary method of interaction and grants the skill the capability to run processes on the host system.
[DYNAMIC_EXECUTION]: The core logic of the skill is executed from a path defined by the environment variable $WIKIPEDIA_MCP_SCRIPT. Loading and executing a script from a path determined at runtime is a dynamic execution pattern, though standard for this specific architecture.

wikipedia-research