The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use rg (ripgrep) to search local files as the first step of its workflow. While this uses the user-supplied topic as a search term, it is a standard implementation for local data retrieval in agentic workflows.
[EXTERNAL_DOWNLOADS]: The skill core functionality relies on fetching external content from URLs discovered during search operations. This is an expected and primary behavior for a web research tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web content, creating a surface for indirect prompt injection attacks where malicious instructions hidden in source pages could attempt to influence the agent.
Ingestion points: Untrusted data enters the context via the WebFetch tool during research steps.
Boundary markers: The skill uses structured templates and a mandatory citation-verification step to help ensure accuracy and structure, though it does not explicitly instruct the agent to ignore commands within source text.
Capability inventory: The agent has the ability to search local files, write to the chat, and save summaries to the filesystem upon user request.
Sanitization: The skill does not define specific sanitization or filtering protocols for external content before it is processed for synthesis.

deep-research