The Agent Skills Directory

[SAFE]: The skill functions as a design assistant, coordinating between the user and specialized agents to produce markdown documentation. No evidence of malicious activity, such as credential theft, network exfiltration, or unauthorized command execution, was found during the analysis.\n- [INDIRECT_PROMPT_INJECTION]: The skill features a surface for indirect prompt injection as it processes untrusted user input and existing documentation to generate new content.\n
Ingestion points: User responses during the brainstorming interview and existing local design files in the gat/overview/ directory.\n
Boundary markers: None present; the agent treats user input as the primary context for document generation without specific delimiters.\n
Capability inventory: The skill uses Write and Edit tools to generate files and the Agent tool to delegate tasks to specialized design and art sub-agents.\n
Sanitization: No explicit input filtering is performed, as the skill is designed to synthesize and document the user's ideas verbatim. Mitigation: Consider wrapping external content in delimiters with explicit instructions for the agent to ignore any embedded directives.

gat-brainstorm