Coordinate multi-model critiques of research hypotheses and designs.
npx skills add https://github.com/sheehe/coase --skill idea-critic
Please help me install this Agent Skill. Skill: idea-critic Source: https://github.com/sheehe/coase/tree/main/resources/plugins/coase-builtin/skills/idea-critic Command: npx skills add https://github.com/sheehe/coase --skill idea-critic
Coordinates independent, multi-model critiques of research hypotheses, designs, and results to improve robustness and insight.
Coordinate independent, multi-model critiques of the given hypothesis set using all selected providers and return per-model analyses with a final consensus summary.
High-intent search queries and answers about installing and using this skill.
To run multi-model critiques on a research hypothesis, configure a diverse panel of providers to perform parallel evaluations. The system coordinates independent critiques and returns per-model analyses alongside a final consensus and divergence summary.
Multimodel consensus evaluation for research design is a technique where independent providers critique a study's methodology. It generates per-model outputs with preserved raw responses and aggregates them into a structured summary highlighting agreement and divergence.
Yes, you can use human-in-the-loop review during multimodel design critique. The system supports configurable critic providers and human-in-the-loop interactions to balance provider bias and comprehensiveness across hypothesis evaluation and result critique.
Available modes for research result critique include discuss, score, design-critique, and result-critique. These modes preserve per-model outputs and generate a unified consensus summary to improve research robustness and insight.
You should use multimodel critiques instead of a single model when you need to balance bias and improve robustness across hypothesis evaluation. Coordinating a diverse panel of providers ensures structured aggregation with consensus and divergence summaries.
ethanhqExecute one-shot or batch tasks using provider subagents.
ManfronEnricoGenerate structured, testable hypotheses and experimental plans from observations and data.
xiechyGenerate testable hypotheses with mechanistic explanations and experimental designs.
haberlahSynthesize consensus, divergence, and recommendations from multi-voice archetype panels.
