The Agent Skills Directory

[COMMAND_EXECUTION]: The agent uses a custom CLI tool, oma, to perform memory operations (oma memory recall/save) and to delegate tasks to sub-agents (oma spawn). These actions are core to the skill's stated purpose as an orchestrator.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided task descriptions and interpolates them into commands for sub-agents, which is an inherent risk factor for orchestrator agents.
Ingestion points: User input for complex coding tasks is processed via the oma spawn protocol defined in SKILL.md.
Boundary markers: The skill suggests wrapping task descriptions in double quotes but lacks explicit instructions for sub-agents to disregard potential instructions embedded within the user data.
Capability inventory: The agent can execute shell commands through the oma CLI to manage system state and delegate work to other models.
Sanitization: No validation or sanitization of the <detailed_task_description> variable is performed before it is executed as part of a shell command.

Sisyphus