mguide

What problem does it solve? Researchers using Mechanist should not have to learn parameter syntax, master writing a compliant task.md, or memorize which skills exist. This Skill acts as a single conversational front door that routes any request to the right place. ## Core Features & Use Cases - Experiment setup (Track A): Works out research requirements through conversation, settles the behavior-source and mechanism parameter axes, writes a compliant task.md into the project root, and submits the /auto run after user confirmation. - Literature search (Track B): Hands explicit search requests directly to /msearch for multi-source retrieval and synthesis. - Field history (Track C): Routes survey and evolution-of-a-field requests to /mhistory for a timeline-based article. - Use Case: A user says "verify whether small models track false beliefs worse than facts." The Skill infers the parameter axes, probes the disk for models and GPUs, asks only what cannot be inferred, writes task.md, and launches the run. ## Quick Start Tell the assistant what you want to study, look up, or survey in plain language and it will set up the experiment or fetch the literature for you.