mise-configuration_skill

Overview

This skill configures a project environment using mise [env] as the single source of truth. It centralizes environment variables, auto-creates Python virtual environments, and supports hub‑spoke layouts for monorepos so each domain can inherit or override settings. The patterns are backward compatible: scripts use normal env fallbacks and work whether mise is installed or not.

How this skill works

Place all configurable values under the .mise.toml [env] section and use standard env access patterns in code (e.g., os.environ.get(...), process.env, ${VAR:-default}). mise will auto-load those values when the shell runs mise activate, create venvs per workspace rules, and apply templating, file loading, and redaction directives. Hub (root) mise.toml supplies shared [tools] and [env]; subfolder spokes extend or override env and declare domain tasks.

When to use it

• You need a centralized environment variables SSoT across a repo or monorepo.
• You want automatic Python venv creation and activation at workspace root or per-project.
• You have a multi-domain project and want hub-spoke task delegation to keep root mise.toml lean.
• You must support multi-account GitHub tokens per directory and ensure tools pick the right env var.
• You need templated, lazy, or tool-aware env values (files, commands, computed values).

Best practices

• Define all configurable values in .mise.toml [env] and keep defaults in code using env.get patterns.
• Pin [tools] at the hub and let spokes inherit to avoid version drift.
• Hoist dev dependencies to the workspace root pyproject.toml for consistent dev environments.
• Delegate domain-specific tasks to spoke mise.toml files when root exceeds ~50 lines.
• Use _.file, _.source, and _.path directives for structured env loading and PATH extension.

Example use cases

• Monorepo: hub mise.toml defines python venv and orchestration; packages/* spokes define build/test tasks.
• ML project: experiments spokes declare EPOCHS and LEARNING_RATE while hub manages python/tool versions.
• Infra repo: root manages terraform and kubectl versions; env tokens and secrets are loaded per-spoke.
• Multi-account GitHub: per-directory GH_TOKEN and GITHUB_TOKEN loaded from secure token files or vault.
• Local-first workflows: scripts use env fallbacks so CI or developers without mise still run reliably.