The Agent Skills Directory

[SAFE]: The skill provides architectural templates and TypeScript code for agent orchestration, including sequential, supervisor, and parallel execution patterns.- [INDIRECT_PROMPT_INJECTION]: The framework implements a multi-agent communication architecture where the output of one agent is used as input for another (e.g., in the SupervisorOrchestrator and fanOutFanIn methods). This is the primary intended function of the skill; however, users should be aware that untrusted data processed by one agent could influence the behavior of downstream agents if appropriate boundary markers or sanitization are not implemented by the developer.- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill utilizes a SharedMemory class which connects to a Redis instance via the REDIS_URL environment variable. This follows standard security practices for managing connection strings and does not involve hardcoded credentials or unauthorized network communication.

ai-agent-orchestrator