Development

Prerequisites

Install development tools:

cd ~/repos/mise-completions-sync
mise install

This installs:

• rust - For building the project
• uv - For running Python scripts
• prek - For pre-commit hooks

Tasks

The project uses mise tasks for common operations. Run mise tasks to see all available tasks.

Task	Description
mise run build	Build the release binary
mise run test	Run tests
mise run lint	Run clippy lints
mise run format	Format code
mise run install-dev	Install locally for testing
mise run install-hooks	Install pre-commit hooks
mise run generate-registry	Generate registry.toml from mise's registry
mise run validate-registry	Validate registry entries against installed tools
mise run audit-registry	Install every registry tool and verify it (slow)
mise run docs-tools	Generate tools documentation from registry

Adding Tools to the Registry

If a tool you use isn't in the registry:

1. Check if the tool supports completions (usually tool completion --help)
2. Add an entry to registry.toml
3. Run mise run validate-registry to verify it works
4. Submit a PR

Registry Entry Format

Most tools follow one of the shared conventions in [patterns], where {} stands in for the tool name:

[tools]
mytool = "standard"      # {} completion <shell>

A tool that doesn't fit a pattern spells out each shell. Omitting a shell means the tool doesn't support it:

[tools]
othertool = { zsh = "othertool completions --shell zsh", bash = "othertool completions --shell bash" }

Optional fields cover the awkward cases: requires names a helper binary the command needs on PATH, provided_by marks a binary that ships inside another mise tool, and bundled means the tool ships completion files rather than generating them (each shell's value is then the filename to look for). See How It Works for details.

Registry Audit

mise run validate-registry only checks tools that happen to be installed on the machine running it, so an entry that is wrong for everyone still passes. The Registry Audit workflow runs weekly, installs every tool in the registry and verifies each command, and opens an issue when something is broken. Tools that won't install on the runner are reported separately — that says nothing about whether the entry is correct.

Run it locally with mise run audit-registry. It installs every tool in the registry, so expect it to be slow and to leave a lot behind.

A few tools need a working environment — not just an install — before they will print a completion script, so the audit can't judge them. kubeseal wants a reachable cluster config even to emit a static file. Those entries carry audit_skip with a reason and are reported as unverified rather than broken:

kubeseal = { audit_skip = "needs a reachable cluster config even to print completions", zsh = "kubeseal completion zsh" }

Use it sparingly: it silences the only check that would catch the entry going stale. mise run validate-registry still tests these normally, since a developer running it locally usually does have the environment.