omni

You can reduce token usage by applying opt-in formatting that compresses prose while strictly preserving code blocks, tool calls, file paths, and structured data. This ensures communication remains fully accurate but optimized for tight token budgets.

Headless mode for agent-to-agent communication provides a raw, unstyled output channel activated via a specific modifier. It facilitates deterministic, token-efficient data exchange between sub-agents in workflow pipelines.

Formatting AI outputs for token-budget pipelines involves using the Skill as a sub-agent to eliminate fluff and echoing. It enforces deterministic notation conventions and maintains zero compression for structured data to ensure pipeline reliability.

Concise response compression does not affect code blocks and file paths. The system enforces strict non-compression for code, tool calls, file paths, URLs, and structured data to preserve full technical accuracy.

You should use safety pauses for ambiguous actions when a workflow involves destructive steps or unclear operations. The system provides failsafe clarity suspension to prevent errors during these critical, high-risk execution phases.