Setup

---
description: "Verify the ruff-format hook's runtime prerequisites and configuration for this repository. Use when: 'set up ruff-format', 'configure ruff-format', 'is ruff-format working', formatting silently isn't happening, or the hook reported a missing prerequisite. Actions: check (read-only verification, default) | apply (resolve what check found). Re-runnable and safe."
argument-hint: "check | apply [install-ruff]"
user-invocable: true
disable-model-invocation: true
---

## Purpose

Thin check-centric setup per the uniform setup contract (`docs/PLUGIN-PHILOSOPHY.md`
"Setup is explicit and repeatable" in the marketplace repository): `check` inspects and
reports, `apply` resolves. This plugin owns no consumer-project configuration. Rules come
from the repository's own Ruff config, and the only tunable is the native `userConfig`
toggle, so `apply` is guidance-and-verify, with exactly one write path: the explicitly
invoked `apply install-ruff` install into the repo's existing managed environment
described below.

Action routing: no argument or `check` runs the check; `apply` runs the check first, then
remediation; `apply install-ruff` additionally authorizes the consumer-repo install
described below. All are non-interactive. Never prompt when the action is given.

## `check` (read-only)

The hook script (`${CLAUDE_PLUGIN_ROOT}/hooks/ruff-format.sh`) is the single source of
truth for what it requires and how it resolves things.

**Read it first.** Probe what it actually does, don't recite this file. Then run each probe via
Bash and report a PASS/FAIL/INFO table with one remediation line per FAIL. Do not modify anything.

When the plugin's toggle is disabled, every prerequisite absence downgrades from FAIL to
INFO. The hook exits through its enabled-gate before probing anything, so a deliberately
disabled plugin is not broken. Report the probes informationally and note that re-enabling
restores the FAIL semantics.

1. **Bash version.** Check against the hook's documented floor (README Requirements),
   noting any features the hook degrades without (for example telemetry's `EPOCHREALTIME`,
   Bash 5.0+).
2. **`jq`.** `command -v jq`. FAIL if absent: the hook then skips with a visible
   once-per-session notice instead of formatting.
3. **Ruff binary.** Resolve it exactly the way the hook's resolution code does: its
   repo-managed virtual-environment walk (the exact `.venv` interpreter paths it tests for
   the current platform, walking up from the edited file toward the repo root) and then
   `PATH`. Test only what the hook tests. A binary the hook would not accept must not PASS
   here. FAIL when nothing the hook would resolve is present while a Ruff config governs the
   repo; the hook then emits a visible once-per-session skip notice instead of formatting.
4. **Consumer Ruff config.** Mirror the hook's opt-in walk: it stops at the FIRST
   (closest) governing config found walking from the edited file's directory up to the repo
   root, honoring Ruff's own same-directory precedence and counting a `pyproject.toml` only
   when it carries a `[tool.ruff]` section or any `[tool.ruff.*]` subtable such as
   `[tool.ruff.lint]`. Read the hook for the exact names and the section test; its test is
   the authority. Report the governing config the walk discovers, or INFO that none exists.
   Absence is the opt-out by design, so the plugin is inert (INFO, not FAIL), matching the
   README's "ships no rules of its own" stance.
5. **Hook toggle.** Report the effective `ruff_format_enabled` value:
   `${user_config.ruff_format_enabled}` (unexpanded or empty means default `true`).
6. **Hook registration.** INFO: confirm the plugin is enabled for this project
   (`/plugin` → Installed) rather than parsing settings files.

## `apply` (idempotent)

Run `check`, then for each FAIL offer the resolution. Never install anything without the
consumer's explicit go-ahead in the invocation. `apply install-ruff` installs Ruff **only
into a managed Python environment the repo already uses**, never by creating one and never
globally, mirroring how the hook resolves the binary. Resolve the target from what the repo
already declares:

Principles, in order. They decide every case, whatever the tool:

1. **Identify the repo's dependency manager from its own markers.** A lockfile or a
   `pyproject.toml` tool section (uv, Poetry, Pipenv, PDM, Hatch, …). Recognize the tool
   from what the repo declares; don't assume from an enumerated list.
2. **Record through the manager, never around it.** A managed project gets Ruff via that
   tool's own dev-dependency add command (e.g. `uv add --dev ruff`,
   `poetry add --group dev ruff`, `pipenv install --dev ruff`, `pdm add -d ruff`) so the
   manifest and lockfile record it. A bare `pip install` into its environment is state
   the tool's next sync or clean silently removes.
3. **Never create or mutate an environment.** When no environment exists yet, use the
   tool's record-only mode when it has one (e.g. `uv add --dev ruff --no-sync`. Plain
   `uv add` syncs and would create `.venv`) and hand the sync/install step to the
   consumer as their own command; when the tool's add command cannot avoid
   creating/instantiating an environment, don't run it. Give it as guidance instead.
4. **Only where the hook resolves.** The hook resolves repo-ancestor `.venv` interpreters
   or `PATH`, nothing else. Tools that default their environment to a cache directory
   (Poetry, Pipenv, and any similar) must have an in-project environment confirmed first
   (the tool's own config/env answers, e.g. `poetry config virtualenvs.in-project`,
   `PIPENV_VENV_IN_PROJECT`); otherwise guide (enable in-project mode + recreate, or put
   `ruff` on `PATH`) rather than installing somewhere the hook never looks.
5. **Bare `pip install` only into a plain existing `.venv`** with no manager markers of
   any kind. State the change and target environment before running.
6. **Ambiguity stops.** No environment plus no recognized manager, or conflicting
   signals → guidance only, anchored on the README's astral install URL
   (`https://docs.astral.sh/ruff/installation/`), matching the hook's own skip-notice
   text.

After ANY remediation, re-run the relevant `check` probe and report its actual result.
Never claim resolved on the install command's exit code alone. For everything else `apply`
only points:

- missing `jq` / Bash: platform install instructions from the README Requirements section;
  this skill never installs system packages.
- toggle off: direct to `/plugin configure ruff-format` (interactive, any
  time). Headless: rerun the install with the new value,
  `claude plugin install ruff-format@<marketplace> -s <scope> --config ruff_format_enabled=true`
  (repeatable per key). Against an already-installed plugin it prints `already installed` **and
  still writes the value**, verified on Claude Code 2.1.240 (a non-sensitive option at `user`
  scope: a non-default value written to an installed plugin, then restored). The short-circuit is
  about the install, not the config write. Re-verify before relying on it outside those
  conditions. A `sensitive` option, or `project`/`local` scope, were not covered. Do **not**
  uninstall to reconfigure: uninstalling drops this plugin's entire stored `pluginConfigs` entry,
  resetting every option in the README's Options reference table to its manifest default. `-s`
  defaults to `user`, so pass the scope `claude plugin list` reports for this plugin, and run from
  that project's directory for a `project`/`local` scope, or the write lands at a scope that does
  not load. This skill never writes user settings or `pluginConfigs`.
  Afterwards, keep the two claims apart. The write is issued and the stored value is what you
  passed; the RUNNING session's behavior is not. The rendered `${user_config.*}` is injected at
  skill load and each hook receives its `CLAUDE_PLUGIN_OPTION_*` from an environment fixed at
  session start, so a same-session `check` still reports the OLD value. Reporting that as a
  failed write would be wrong. Verify the effective value by rerunning `check` in a **fresh
  session**, and never claim an unobserved change.
- no Ruff config: offer to create a minimal Ruff config in the repository root only when
  explicitly asked. The plugin imposes no rules of its own.

Re-running `apply` after everything passes changes nothing and reports "already configured".

## What this skill does NOT do

- Run the formatter. Editing any `.py`/`.pyi` file exercises the hook end-to-end.
- Write the plugin cache, Claude Code user settings, or `pluginConfigs`.
- Create a virtual environment, install Ruff globally, or install outside a managed
  environment the repo already uses.
- Download or execute tools during `check`; network use happens only in an explicitly
  requested `apply install-ruff` inside the consumer repository.