nyldn/claude-octopus ? reverse-engineered prompt

Build me a Claude Code workflow that can run the same research, design, or coding task through multiple AI models and then show me where they agree and where they disagree before I trust the result.

I want the default path to stay simple when Claude alone is enough, but I also want an easy way to escalate into a multi model review or council when I need stronger confidence. Make it feel like a real developer tool I can use from slash commands, with clear phases for discovering the problem, defining the plan, developing the work, and delivering the final answer. If possible, include support for saving context between sessions, and make it easy to plug in extra providers when they are available.

Please set it up so the system is careful about quality, can flag conflicts between models, and can give me a final recommendation after comparing responses. If you need to look up current docs online, go ahead.