MiroFish Guide
A community skill for operating MiroFish: prepare simulation inputs, diagnose pipeline failures, compare model routes, and audit reports against runtime evidence.
What is MiroFish Guide?
MiroFish Guide gives AI coding agents a reusable reference for working with the MiroFish simulation engine. Maintained by MOZARTINOS, it is a community companion to the upstream project. The repository supplies a SKILL.md entry point and focused reference files for Claude Code, OpenClaw, Codex-style workflows, and similar agents. The guide covers source preparation, prompt design, graph extraction, persona generation, model and proxy compatibility, runtime troubleshooting, and output evaluation. Its operator workflow starts with one scenario and a small pilot, checks generated artifacts, and records changes before expanding a run. Report auditing is a central part of the workflow. It asks operators to trace conclusions back to action logs, simulation databases, and report-agent tool use, and to distinguish observations from unsupported claims. Use it to investigate failures and document repeatable experiments. Simulation reports should be treated as exploratory qualitative narratives rather than validated forecasts. Open the repository and point your agent at SKILL.md, then use the references relevant to your current stage. Operating simulations requires the separate MiroFish engine and its configured services.
Best for
Technical MiroFish operators and coding-agent users who want repeatable simulation setup, artifact-based debugging, model-route comparisons, and report evaluation.
Not for
Users seeking a hosted simulation product, a beginner course, or validated forecasts of real-world behavior. The guide does not replace the MiroFish engine.