ci skill

Put Vale in the pipeline

When to use this

Vale runs locally and should now run automatically.

Prerequisites

1. vale ls-config resolves, and a plain vale <path> run works locally. Do not wire up CI for a config that does not resolve — the failure will look like a CI problem.
2. The user has said whether the job should block. Ask if not.

Advisory or blocking

Only error sets a non-zero exit code. Warnings and suggestions exit 0, so a job wired up naively passes no matter how much it reports.

• Advisory — Vale reports, the job stays green. The right way to adopt Vale on a corpus that has never had it.
• Blocking — the rules the team cares about are set to error in the config, and the job fails on those.

Say which one you are setting up, in those words. "It runs in CI" tells nobody whether it can stop a merge.

GitHub Actions

- uses: vale-cli/vale-action@reviewdog
  with:
    files: docs

Alerts arrive as review comments on the pull request, which is where a writer will see them.

pre-commit

repos:
  - repo: https://github.com/vale-cli/vale
    rev: v3.17.0
    hooks:
      - id: vale

Pin rev to a release. Hooks that float break at the worst time.

Any other runner

vale --no-global --output=line docs/

--output=line gives path:line:col:rule:message, which most CI systems can turn into annotations. --no-global keeps a config on the machine from changing the result.

If the runner has no cached StylesPath, vale sync has to run first — and that means a style updating upstream can fail a build nobody changed. Pin Packages to a release URL rather than a moving reference.

Do not

• Do not add --no-exit to make a failing job pass. If those alerts should not block, change the levels in the config, where the decision is visible to the team.
• Do not raise MinAlertLevel in CI only. A config that reports different things locally and in CI is how people stop trusting the tool.