Actions reference

An action is pure data: a parameter declaration plus a shell template. Nothing in the execution path knows about actions — a flow is compiled into a plain script when you save the job, which is why remote runners need no knowledge of them.

There are 33 built-in actions in nine categories.

General

Action	ID	What it does
Run command	shell.run	Runs a shell command in the working directory. The general-purpose action.
Print message	shell.echo	Writes a line into the build log. Useful for marking progress.
Wait	shell.sleep	Sleeps for a number of seconds — typically waiting for a service to come up.
Set environment variable	shell.env	Sets a variable for subsequent steps by appending to $WALIS_ENV_FILE.
Set output	shell.outputs	Writes name=value as this step's output, for later steps to reference as ${{ steps.ID.NAME }}.
Change directory	shell.cd	Switches directory; later steps run there too.

Files

Action	ID	What it does
Create directory	file.mkdir	Does not fail if the directory already exists.
Copy file	file.copy	Files or directories.
Move file	file.move	Move or rename.
Delete file	file.remove	Does not fail if the path is absent.
Zip	file.zip	Compresses a directory or file into a zip.

Git

Action	ID	What it does
Git clone	git.clone	Optional branch and depth.
Git checkout	git.checkout	Switch to a branch or commit.
Git pull	git.pull	Fetch and merge the latest commits.

Node

Action	ID	What it does
Install dependencies	node.install	Installs from the lock file.
Run npm script	node.script	Runs a script from package.json.
Run npx command	node.npx	Runs a command a package provides.

.NET

Action	ID	What it does
dotnet restore	dotnet.restore	Restores NuGet dependencies.
dotnet build	dotnet.build	Builds a project or solution.
dotnet test	dotnet.test	Runs test projects.
dotnet publish	dotnet.publish	Produces publish output.

Containers

Action	ID	What it does
Docker build	docker.build	Builds an image.
Docker push	docker.push	Pushes an image to a registry.
Docker compose	docker.compose	Brings services up or updates them.
Run in a container	container.run	Mounts the working directory into an image and runs a script inside it — the equivalent of Jenkins' agent { docker }.

Deployment

Action	ID	What it does
SSH remote command	deploy.ssh	Runs a command on a remote host — restarting a service, running a deploy script.
rsync	deploy.rsync	Syncs a directory to a remote host.
scp upload	deploy.scp	Copies files to a remote host.
HTTP request	deploy.http	Calls an endpoint — triggering a deploy, posting a notification.

Data

Action	ID	What it does
HTTP request	http.request	Sends a request with curl and publishes status and body as this step's outputs. A 4xx or 5xx does not fail the step — that is a decision for a run condition.
Read a JSON field	data.json-get	Pulls a field out of JSON (usually a previous http.request body) and publishes it as value. Path is dot-separated, arrays by number. Needs python3 or node on the runner.

These two declare their outputs, so they appear in the Insert an output of an earlier step menu. Together with a run condition they are how a flow calls an API, reads a field and branches on the answer.

Checks

Action	ID	What it does
File exists	check.file	Fails the build when the file is missing.
Command available	check.command	Fails the build when the command is not on PATH.

Per-step settings

Every step in a flow carries these settings, independent of which action it uses:

• Enable this step — off means it is skipped without failing the build.
• Continue on failure — the build carries on past a non-zero exit.
• Depends on — which steps must finish first. Steps that depend on nothing start together; see the flow is a graph.
• Identifier — a short unique name so later steps can reference this one's outputs.
• Run condition (expression) — evaluated against earlier steps' outputs at run time. False means skipped, not failed.
• Run when — only if everything before succeeded (the default), always, or only after a failure. This is Jenkins' post { always / failure }: a cleanup or notification step still runs after something has gone wrong, while ordinary steps are marked Skipped. The build still counts as failed.
• Retries — 0 to 10 further attempts. Each attempt is a fresh shell, with the earlier output kept above it. Retries never extend the job timeout.

Script mode has none of these: a single compiled script has nowhere to put them.

Custom actions

A custom action is the same shape as a built-in one, stored as JSON in the database. You declare parameters and write a shell template; the template language is documented in the project's Docs/actions.md. Custom actions travel with export and import.

If you need behaviour rather than a template — logic, branching, calling an API — write a plugin instead.

Action packs

Manage actions has an Action packs section with a Reload button. Packs are shared action definitions kept in a repository — the public one is walisayu/walis-actions — pulled in as files rather than installed as code. They are still just parameter declarations and shell templates, so the same thing holds as for any action: you can read what will run before you use it.