reader: someone installing or running meow-mise on Pi answers: what meow-mise provides, what it adds to a session and how to install it kind: reference describes: [@meowshed/[email protected]]

@meowshed/meow-mise

Reports the tasks mise resolves in a repository, with where each came from and what stops a stage running it unattended, binds the stages to them, and never grants trust, answers a prompt or writes a file. Distributed as a Pi package.

Install it

pi install npm:@meowshed/meow-mise

The paw executable ships with @meowshed/meow-core alone (REQ-4612): this package's launchers run their subcommands through it, so install the core package beside this one for the binary-backed checks. Without it, every check reports unrun and lets the command through.

pi install npm:@meowshed/meow-core   # carries the binary the launchers run

Installing from this repository's checkout is the development install: it loads the package in place and runs no lifecycle script.

Bind your stages to your tasks

meow-mise bind prints a [stages] table for you to paste into .meowpaw/profile.toml, and never writes the profile itself. Commit what it prints, so the stages run from the declaration the repository holds:

[stages]
test = "mise run --force test"
# lint: task lint is blocked: hidden

It binds a stage only to the task of exactly its name, so a task named tests or unit is never bound to test, and only a task that carries no block. Every binding runs the task with --force, so a task mise would skip as fresh runs, and a passing stage is never a skip. Each stage it leaves unbound is a comment with the reason.

meow-mise check reads your profile's [stages], finds each mise run <task> in a stage's command, including each part of a chain joined by &&, || or ;, and reports every block on that task, a task it can't find, and a task that can skip as fresh run without --force. It exits 0 on no finding, 1 on a finding and 3 when it can't read the profile or mise, so you can run it in your own gate.

check reads the profile, so it prints the profile's state, profile: absent, profile: unparseable or profile: parsed, and unknown key: <path> for each key no unit reads, which changes no exit status.

What it reports instead of a list

status exits 0 when it reports the tasks in every directory and 3 when it can't in one, and no command ever reports a state it couldn't read as an empty list. bind and check read the root alone, because they read the root profile's stages:

| Line | Means | | ------------------------------------- | ------------------------------------------------------------------------- | | unresolved: not a mise repository | Git tracks no mise configuration file or task directory; mise didn't run | | unresolved: mise not found | mise isn't on PATH | | unresolved: untrusted <directory> | mise needs the configuration trusted; read it, then run the command shown | | unresolved: environment: ... | Your mise is older than 2026.9.11, the version the unit was tested on | | unresolved: meow-mise defect: ... | mise rejected a flag this unit passes; report it | | unresolved: unrecognised shape: ... | mise printed a listing this unit can't read | | unresolved: mise failed ... | mise failed another way; its last lines follow |

Listing the tasks runs mise over your configuration, and on a trusted configuration mise evaluates its templates, which can run code. The report names each committed configuration file whose templates call exec.

What it costs you

The skill's description stays in context on every turn, within the 380 characters the unit's budget states. The program is a native binary shipped inside the unit, and needs mise and git on the machine. On a machine the unit carries no binary for, it reports the repository as unresolved and exits 3.

Read a subcommand's help

meow-mise --help lists the subcommands, and meow-mise <subcommand> --help names what it determines, what you read to reach the same answer without the tool, and the first word of each kind of line it prints. A test holds each subcommand's output to those words.