The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill serves as an orchestration layer that ingests external template data and user context to drive sub-agent behavior, creating a surface for indirect prompt injection.\n
Ingestion points: Reads workflow templates from ~/.maestro/templates/workflows/, accepts user-provided context via --context flags, and processes artifacts produced by sub-agents.\n
Boundary markers: The skill uses a defined state machine and CSV schema for data transfer but does not explicitly document sanitization of interpolated variables before they are placed into sub-agent instructions.\n
Capability inventory: The skill utilizes spawn_agents_on_csv to execute tasks, and the coordinator maintains Bash, Read, Write, and Edit tools to manage the workflow lifecycle.\n
Sanitization: No specific escaping or validation logic is detailed for the {variable} or {N-xxx.field} resolution process.\n- [COMMAND_EXECUTION]: The skill generates and executes skill calls dynamically based on the contents of the workflow templates. While the coordinator delegates execution to sub-agents via spawn_agents_on_csv, the logic facilitates the execution of any skill call defined within the ingested template files.

maestro-player