The Agent Skills Directory

[SAFE]: Multi-agent orchestration logic. The skill acts as an orchestrator for several sub-agents (Planner, Architect, Critic) using a consensus-based iterative loop to ensure plans are vetted before handoff to execution.- [SAFE]: Input validation gate. Implements a logic gate that prevents the execution of vague or underspecified commands, requiring a specific level of detail (e.g., file paths, issue numbers, or explicit code blocks) before proceeding, which serves as a security and efficiency filter.- [SAFE]: Local context persistence. Reads and writes task-related metadata to the '.omx/context/' directory to maintain state between workflow steps, adhering to local environment constraints.- [SAFE]: Review loop mitigation. The inclusion of an Architect role for soundness checks and a Critic role for quality enforcement provides a built-in safety mechanism against the generation of unsafe or low-quality plans.

ralplan