Ln 310 Multi Agent Validator

This is a heavyweight L2 coordinator that runs multi-phase validation through an evaluation platform with mandatory research workers, parallel evidence gathering, and sequential merge and refinement steps. It operates in two modes: story validation and plan review. Every run requires external research (official docs, MCP reference, web), launches read-only evidence lanes in parallel, then sequentially generates docs, applies repairs, merges findings, and runs refinement before final approval. The workflow is rigid by design: ten phases with explicit runtime checkpoints, worker delegation via the Skill tool, and agent health checks that distinguish transport failures from actual domain findings. If you need a lightweight quick check, this is not it. If you need auditable multi-agent validation with research requirements and phase accountability, this enforces the full ceremony.

Install to Claude Code

npx -y skills add levnikolaevich/claude-code-skills --skill ln-310-multi-agent-validator --agent claude-code