The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the run_command tool to execute the agy CLI, which is the primary mechanism for dispatching background worker tasks.
[DYNAMIC_EXECUTION]: Shell commands for the agy CLI are dynamically assembled at runtime, incorporating model selection, effort levels, and task-specific prompts.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests user-provided tasks and interpolates them into worker prompts.
Ingestion points: User input provided via the router command is processed and passed to background workers.
Boundary markers: The skill uses the -p="..." flag in the CLI to encapsulate prompts, but does not specify explicit escape sequences for untrusted content within those strings.
Capability inventory: The skill uses run_command (for CLI execution), manage_task (for lifecycle management), and reads/writes temporary files in /tmp.
Sanitization: No specific sanitization or filtering logic is defined for the user-supplied task descriptions before they are passed to the agy command.

router