The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill orchestrates complex multi-agent research by ingesting findings from various sub-agents into subsequent prompts for synthesis and refinement, creating a vulnerability surface where malicious data could influence agent behavior.
Ingestion points: The instructions.md file defines templates for Generate, Aggregate, and Refine operations that interpolate PARENT NODE CONTENT and NODE CONTENT directly into the agent's prompts.
Boundary markers: The templates use standard text headers to separate research findings but do not implement formal delimiters or specific instructions for the agent to ignore potentially malicious embedded commands within those findings.
Capability inventory: The skill uses the Task Tool for deploying sub-agents and the Read/Write tools for persisting the graph state to the file system.
Sanitization: There is no evidence of content validation, escaping, or sanitization of the research data received from external sub-agents before it is re-interpolated into the controller's context.
[EXTERNAL_DOWNLOADS]: The skill includes a conceptual reference link to the official Graph of Thoughts repository hosted by SPCL at ETH Zürich as a foundation for its methodology.

got-controller