The Agent Skills Directory

[SAFE]: The skill is a standard research assistant that performs structured information gathering and synthesis. All identified behaviors align with its stated purpose of scientific literature review.
[EXTERNAL_DOWNLOADS]: The skill utilizes the mcp_base_search_literature and search_web tools to retrieve scientific papers and public web content. These activities target reputable academic sources and well-known search engines.
[PROMPT_INJECTION]: While the skill ingests untrusted external data during its search phase, the risk of indirect prompt injection is minimal given the lack of dangerous capabilities (such as arbitrary code execution) that could be leveraged by such an attack.
[COMMAND_EXECUTION]: The skill instructs the agent to use tool-calling capabilities for filesystem operations within a controlled research directory (mcp_base_create_research_dir), which is a safe and standard implementation for generating research outputs.

general-deep-research