What problem does it solve?
This Skill provides a structured, probabilistic assessment of a codebase or project by orchestrating multiple raters and adversaries to produce a final rating, helping teams gauge quality quickly.
Core Features & Use Cases
- Parallel multi-agent evaluation: spawns N raters to sample diverse aspects of the repository and produce individual ratings.
- Adversarial refinement: for each rater, K subagents generate adjusted reports that challenge the original assessment.
- Improve mode: optional flag that collects actionable improvements from subagents and includes them in the final synthesis.
- Final synthesis: aggregates surviving ratings into a single mean rating and provides a concise synthesis of strengths and gaps.
Quick Start
Invoke /rate in chat to obtain a probabilistic rating for your codebase.