marimo-pair

The marimo-pair skill enables direct interaction with a running marimo notebook's kernel, allowing users to execute code, create and edit cells, and manage notebooks programmatically. It is designed for scenarios where a user wants to either start a marimo notebook or continue work in an active session. By working inside the kernel, the skill ensures that all code runs in the live notebook environment, reflecting changes instantly and supporting a seamless development workflow without needing separate local Python environments.

Key features of marimo-pair include reading and inspecting existing notebook cells, creating new cells, running code (both single-line and multi-line), installing packages, and observing the notebook's reactive graph. The skill leverages bundled Bash scripts such as `discover-servers.sh` and `execute-code.sh` for server discovery and code execution, with optional use of the MCP interface for similar operations. It handles execution over HTTP to the notebook kernel, so errors are tied to the code itself rather than the local Python setup. It also supports multi-line code execution and targeting specific sessions or server URLs, providing flexible programmatic control over notebook operations.

The skill is particularly useful for developers, data scientists, or analysts who frequently work with marimo notebooks and require programmatic automation of code execution and notebook management. It is suitable for live pair programming scenarios where users can see results in real-time, as well as for automating repetitive tasks within notebooks. By enabling cell-level control and integration with existing project tooling, marimo-pair helps users maintain efficient, interactive, and reproducible workflows in Python-based data analysis, machine learning experiments, and computational research environments.