The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill retrieves protocol guidelines and output schemas from the author's GitHub repository (csark0812/toolbox). These are standard vendor resources used to maintain consistency across the agent's toolbox.
[PROMPT_INJECTION]: The skill is designed to ingest and analyze external data from the web and local file systems to verify hypotheses. While this creates a surface for indirect prompt injection, the instructions prioritize evidence extraction and logical verification, which naturally limits the impact of adversarial content embedded in the source material.
Ingestion points: The skill reads "primary material" including web content, documents, and repository code as part of its core loop.
Boundary markers: The skill does not explicitly define XML-style tags for segmenting untrusted data inputs, relying instead on the structured output format to separate evidence from interpretation.
Capability inventory: The skill can invoke other sub-agents and specialized tools for research and diagnosis based on the investigation results.
Sanitization: No explicit sanitization of external content is performed; however, the requirement for citable evidence acts as a manual verification step for the user.

investigate