> For the complete documentation index, see [llms.txt](https://developer.harness.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.harness.io/harness-platform/3.0/harness-platform-resources/overview/strategy/conditional-execution.md).

# Conditional execution

Skip a stage or step at runtime by setting a condition that must evaluate to true for it to run.

**Conditional execution** lets you skip a stage or step based on a condition evaluated at runtime. If the condition evaluates to `false`, Harness skips it and marks it as not executed; execution continues from the next stage or step.

Use conditional execution to:

* Run a deploy stage only when the pipeline has succeeded up to that point
* Always run a cleanup stage regardless of prior failures
* Skip a stage when a pipeline input does not match a target value

## What you will learn

* How to skip a stage or step at runtime using the `if` key and a JEXL condition
* What the built-in base conditions (`OnPipelineSuccess`, `OnPipelineFailure`, `Always`) do
* How to combine conditions and gate execution on a pipeline input value
* How skipped stages behave and what happens to their outputs

***

## YAML

Use the `if` key on a stage or step. The condition is a JEXL expression:

```yaml
stages:
  - name: deploy-prod
    id: deploy_prod
    if: <+OnPipelineSuccess>
    steps:
      - name: deploy
        id: deploy
        run:
          script: ./deploy-prod.sh
```

If the expression resolves to `false`, the stage is skipped entirely. No steps run, and no failure is recorded.

***

## Base execution conditions

These built-in JEXL conditions control when a stage runs relative to the pipeline state:

| Condition              | Runs when                                        |
| ---------------------- | ------------------------------------------------ |
| `<+OnPipelineSuccess>` | All previous stages succeeded (default behavior) |
| `<+OnPipelineFailure>` | At least one previous stage failed               |
| `<+Always>`            | Always runs, regardless of prior stage status    |

```yaml
# Run only if prior stages succeeded (same as default)
if: <+OnPipelineSuccess>

# Run a cleanup stage even when previous stages failed
if: <+Always>

# Run a rollback stage only on failure
if: <+OnPipelineFailure>
```

***

## Combine with an expression

You can combine a base condition with an additional JEXL expression using `&&`:

```yaml
# Skip this stage (force-skip using && false)
if: <+OnPipelineSuccess> && false
```

To gate on a pipeline input value, wrap the input expression in single quotes so the resolved value becomes a JEXL string literal:

```yaml
pipeline:
  inputs:
    env:
      type: string
      default: production
  stages:
    - name: deploy-prod
      id: deploy_prod
      if: "'<+inputs.env>' == 'production'"
      steps:
        - name: deploy
          id: deploy
          run:
            script: ./deploy-prod.sh
    - name: deploy-staging
      id: deploy_staging
      if: "'<+inputs.env>' == 'staging'"
      steps:
        - name: deploy
          id: deploy
          run:
            script: ./deploy-staging.sh
```

{% hint style="info" %}
When comparing a pipeline input value, wrap `<+inputs.varName>` in single quotes: `"'<+inputs.env>' == 'production'"`. Without the outer quotes, JEXL treats the resolved value as an identifier rather than a string literal, causing an expression evaluation error.
{% endhint %}

***

## Behavior

* A skipped stage does not count as a failure.
* Downstream stages that depend on a skipped stage's outputs will receive empty values for those outputs.
* Conditional execution is evaluated before the stage starts; it cannot be based on that stage's own steps.

{% hint style="info" %}
In a parallel group, each stage has its own condition. One stage in the group can be skipped while the others still run.
{% endhint %}

***

## Next steps

* Go to [On failure](/harness-platform/3.0/harness-platform-resources/overview/strategy/on-failure.md) to control what happens when a stage fails.
* Go to [Stage inputs & outputs](/harness-platform/3.0/harness-platform-resources/overview/overview/inputs-and-outputs.md) to reference stage outputs in conditions.
* Go to [Pipelines overview](/harness-platform/3.0/harness-platform-resources/overview.md) to review the pipeline model.