agentleFS
saleorsaleor/saleor-dashboard/.claude/skills/saleor-dashboard-github-workflows/SKILL.md

saleor-dashboard-github-workflows

Skill1.1k starsChanged 8 days agoBSD-3-ClauseView on GitHub

Rules for writing and editing GitHub Actions workflows and composite actions in this repository - runner images, action pinning, shell quoting, untrusted interpolation, permissions, and the actionlint gate. Use when creating or changing anything under .github/workflows/ or .github/actions/, when a workflow review comment needs applying, or when a CI job needs adding to the repo.

  • Installs packages

Trending

Less talk lately.about the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 3 mentionsMar 2026: 5 mentionsApr 2026: 0 mentionsMay 2026: 1 mentionJun 2026: 33 mentionsJul 2026: 3 mentionsAug 2026: 0 mentionsSep 2026: 0 mentionsOct 2026: 2 mentionsNov 2026: 0 mentions
Dec 2025Jun 2026Nov 2026
Mentions a month of the saleor/saleor-dashboard repository that our weekly crawl found. There is no adoption data for this file yet.

Who's talking about saleor/saleor-dashboard

The biggest 5 of 50 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 50

Talked about on GitHub (50 times).

On GitHub · 50

45 more on GitHub

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Do not write pnpm run actionlint -- <files>.

    line 42
  2. Very specific:

    If .object.type comes back tag rather than commit (an annotated tag, as pnpm/action-setup uses), dereference it:

    line 97
  3. Distinctive:

    When a tool has no action, download a pinned release and verify its digest rather than piping a script from a branch:

    line 108
  4. Very specific:

    Use ubuntu-24.04. ubuntu-22.04 is deprecated and being retired.

    line 127

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

GitHub Actions in saleor-dashboard

Workflows are the one place where a mistake runs with repository credentials against every pull request. The rules below are what review asks for here, so apply them while writing rather than after the comment arrives.

Non-negotiable: actionlint must pass

actionlint runs shellcheck over every run: block, type-checks ${{ }} expressions, validates runs-on labels, and catches needs/matrix references that do not resolve. Run it before handing work back.

pnpm run actionlint .github/workflows/your-file.yml   # one file
pnpm run actionlint                                   # every workflow

Nothing to install: the script runs the author's official image, pinned by digest in package.json, and Docker is already a prerequisite for pnpm e2e. The image bundles its own shellcheck, so results do not vary with what is on the host.

There is no usable npm distribution — the actionlint and node-actionlint packages are unrelated 2022 forks that predate ubuntu-24.04 as a known runner label, so they would reject workflows that are in fact correct. If you want a native binary for editor integration, brew install actionlint — but the package.json script is what CI runs, so it is the one that decides.

lint-workflows.yml enforces this in CI, but only on the workflow files a pull request touches — the repository predates the check and its older workflows still carry findings. Editing a workflow means inheriting whatever it already reports. Budget for that; do not silence it with # actionlint-disable.

Two gotchas:

  • A run: comment whose first word is shellcheck is parsed as a shellcheck directive and fails the lint. Reword the comment.
  • Do not write pnpm run actionlint -- <files>. pnpm forwards the -- to the script, where it is an end-of-options marker, so the next argument is read as a file name. Pass arguments directly: pnpm run actionlint <files>.

Never interpolate ${{ }} into a run: block

${{ }} is substituted textually before the shell starts, so anything an outside contributor controls — a branch name, a PR title, a workflow_dispatch input — becomes shell source. Pass it through env: instead, where it is only ever a variable value.

# Wrong - the value is spliced into the script
- run: node e2e/run.ts --target=${{ matrix.target }}

# Right
- env:
    E2E_TARGET: ${{ matrix.target }}
  run: node e2e/run.ts --target="$E2E_TARGET"

${{ }} is fine in with:, if:, env: and name: — those are not shell.

Quote every expansion, and build optional arguments as an array

Unquoted $VAR word-splits, so a value containing a space silently becomes two arguments. ${VAR:+--flag "$VAR"} does not solve this: the quotes inside the expansion are not re-evaluated.

# Wrong - a grep value with a space becomes two arguments
run: playwright test ${GREP:+--grep "$GREP"}

# Right
run: |
  args=()
  if [[ -n "$GREP" ]]; then
    args+=( --grep "$GREP" )
  fi
  playwright test "${args[@]}"

Pin actions by commit SHA, with the version in a trailing comment

- uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

A tag can be moved; a SHA cannot. Resolve one with:

gh api repos/actions/checkout/releases/latest -q .tag_name
gh api repos/actions/checkout/git/ref/tags/v7.0.1 -q .object.sha

If .object.type comes back tag rather than commit (an annotated tag, as pnpm/action-setup uses), dereference it:

gh api repos/pnpm/action-setup/git/tags/<sha> -q .object.sha

Start new workflows on current versions. Parts of this repository still pin older majors; copying an existing step forward carries the staleness with it, and review flags it. Check the latest release rather than grepping a neighbour.

When a tool has no action, download a pinned release and verify its digest rather than piping a script from a branch:

env:
  TOOL_VERSION: 1.7.12
  TOOL_SHA256: 8aca8db9...
run: |
  curl -fsSL -o tool.tar.gz "https://github.com/owner/tool/releases/download/v${TOOL_VERSION}/..."
  echo "${TOOL_SHA256}  tool.tar.gz" | sha256sum --check --strict

A container image is the same idea: pin it by digest, and write the registry host out in full. rhysd/actionlint resolves under Docker, which assumes Docker Hub, but not under Podman, which does not guess a namespace it does not know. docker.io/rhysd/actionlint:1.7.12@sha256:... works under both.

Runner images

Use ubuntu-24.04. ubuntu-22.04 is deprecated and being retired; much of this repository still names it, which is drift rather than a pattern to copy. Pin the version instead of ubuntu-latest so a runner rollout cannot change a run's behaviour without a commit.

Permissions

Declare permissions: {} at the top level and grant each job only what it needs. A job that just reads code wants contents: read. Add persist-credentials: false to actions/checkout unless a later step actually pushes.

Prefer an expression to a job

A job costs a runner allocation. Expanding an input into a matrix, or picking between two constants, belongs in an expression:

# Instead of a `resolve-targets` job that echoes to $GITHUB_OUTPUT
matrix:
  target: ${{ fromJSON((inputs.target == 'stable' || inputs.target == 'main') &&
    format('["{0}"]', inputs.target) || '["stable","main"]') }}

GitHub's && and || return operand values, not booleans, which is what makes the ternary-style form work.

Checklist before handing back

  • actionlint clean on every file touched
  • pnpm exec oxfmt --write on the YAML
  • No ${{ }} inside any run: block
  • Every shell expansion quoted; optional arguments built as an array
  • Actions pinned by SHA, on the current release, version in a comment
  • runs-on: ubuntu-24.04
  • Top-level permissions: {} plus a per-job grant
  • concurrency group set, so superseded pushes cancel
  • timeout-minutes set, so a hung job does not burn an hour

Repository specifics

  • Reusable composite actions live in .github/actions/<name>/action.yml.
  • Node and pnpm versions come from package.json (engines, packageManager); use node-version-file: "package.json" and let pnpm/action-setup read packageManager rather than naming versions in the workflow.
  • The E2E suites are e2e/ (local Docker stack, e2e-local.yml) and the frozen e2e-legacy/ (Saleor Cloud). See e2e/README.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.