A library of reusable GitHub Actions workflows and composite actions.
| Name | Kind | Does |
|---|---|---|
| OpenTofu | Reusable workflow | Checks, tests, plans and applies one root module per call, with its var files, in its environment, when a PR touches it; Azure OIDC or secrets; PR comments; approval-gated apply; providers cached across jobs |
| OpenTofu drift | Reusable workflow | Plans one deployment on a schedule, never applying, and reports drift as a GitHub issue per deployment (updated in place, marked resolved when it's gone) |
| Data Factory | Reusable workflows + composite actions | Validate and export a factory's Git folder to an ARM template on every PR (no Publish, no adf_publish), what-if plans per environment, approval-gated deploys with trigger handling; network and compute (runtimes, endpoints) left to IaC; PR comments |
| Synapse | Reusable workflows + composite actions | The same for a Synapse workspace's artifacts (no workspace_publish), deployed with GitHub OIDC; network and compute (pools, runtimes, endpoints) left to IaC; PR comments |
| SQL project | Reusable workflows + composite actions | Build an SDK-style SQL project's dacpac once, DeployReport plans per database with the change list and script in PR comments, additive or source-of-truth deploys, possible data loss blocked by default, approval-gated SqlPackage publish; Entra ID via OIDC |
Reference a reusable workflow from a job:
jobs:
opentofu:
uses: JoshSLawrence/actions/.github/workflows/opentofu.yaml@v0.3.0
permissions:
actions: read
contents: read
id-token: write
pull-requests: write
with:
working-directory: iac/network
var-files: prod.tfvars
apply-environment: prodEach entry's README has its inputs, the setup it needs, and examples. The
Data Factory, Synapse and SQL project composite actions can also be used on
their own;
OpenTofu's are internals of its workflow.
Complete caller workflows are in examples/.
- mise installs every tool, at the versions each project pins.
- No inline scripts in YAML. A
run:step runs one script, where the logic can be read, shellchecked and run locally. A pre-commit hook enforces this. - Easy to follow beats early abstraction.
- Stay DRY with the shared library (
shared/), as long as that doesn't make the code harder to follow. Areas build on it but never on each other, so a change to one area can't break another.
This project is in pre-release (v0.x.x). Breaking changes may occur
between minor versions. Once stable, it will move to v1.0.0.
Releases are tagged vX.Y.Z. Pin to a specific version or use a full commit
SHA (with a # vX.Y.Z comment for Dependabot) for stability.
Workflows refer to their sibling workflows and actions with GitHub's
self-repository syntax (uses: $/...). Whatever ref you pin, everything
therefore comes from the same commit. That syntax needs GitHub Actions
runner 2.336.0 or newer; GitHub-hosted runners always qualify.
This repository is public and available for use in any workflow.
actions/
├── .github/
│ ├── actionlint.yaml # temporary: actionlint doesn't know `$/` yet
│ ├── dependabot.yml
│ ├── scripts/ # this repository's own checks (lint, sync)
│ └── workflows/
│ ├── ci.yaml # lint + end-to-end runs of the workflows
│ ├── datafactory-deploy.yaml
│ ├── datafactory.yaml
│ ├── opentofu-drift.yaml
│ ├── opentofu.yaml
│ ├── sqlproject-deploy.yaml
│ ├── sqlproject.yaml
│ ├── synapse-deploy.yaml
│ └── synapse.yaml
├── arm/
│ └── scripts/ # what Data Factory and Synapse share (arm.sh)
├── datafactory/ # build, deployments, plan, apply, scripts/
├── docs/
│ └── design/ # design documents and their decisions
├── examples/ # caller workflows to copy from
├── mise.toml # tool versions for local hooks and CI lint
├── opentofu/
│ ├── apply/ # composite actions, one per directory
│ ├── azure/
│ ├── checks/
│ ├── drift-report/
│ ├── plan/
│ ├── prepare/
│ ├── provider-cache/
│ ├── README.md
│ └── scripts/ # the logic; actions are thin wrappers
├── shared/ # the library every area builds on
│ ├── pr-comment/ # composite actions every area uses
│ ├── result/
│ ├── scripts/ # common.sh and generic steps
│ └── setup/
├── sqlproject/ # build, deployments, plan, apply, scripts/,
│ # README.md and source-of-truth.md
├── synapse/ # build, deployments, plan, apply, scripts/
└── tests/
├── arm/ # Data Factory and Synapse script tests
├── fixtures/ # what CI runs the workflows against
├── opentofu/ # OpenTofu script tests
└── sqlproject/ # SQL project script and engine tests
mise install
pre-commit install
.github/scripts/lint.shCI runs the same hooks (Lint), plus end-to-end runs of the reusable
workflows against tests/fixtures/. See
CLAUDE.md (also AGENTS.md) for conventions.