STARTER_CHARACTER = 🔬

Ralph Review (Deep)

Validate the beads graph and launch three independent reviews of the ralph artifacts in parallel using different models via opencode run. Each model produces one review file evaluating goal clarity, task decomposition, sequencing, plan completeness, risk and gaps, and feasibility.

For a faster single-model review using an internal subagent, use the ralph-review skill (/ralph-review, $ralph-review) instead. This deep variant trades wall time and cost for cross-model consensus.

Models

Procedure