Plugins

Woodpecker keeps dataset-family behavior in plugins. Plugins register fixes under a stable namespace prefix and may bundle recipes in a package recipes/ directory.

Plugin package	Prefix	Fixes	Recipes	Status
`roocs-woodpecker-atlas-plugin`	`atlas`	2	1	bundled
`woodpecker-cmip6-plugin`	`cmip6`	1	0	bundled
`roocs-woodpecker-cmip6-decadal-plugin`	`cmip6_decadal`	15	1	bundled
`woodpecker-cmip7-plugin`	`cmip7`	3	2	bundled
`woodpecker-xmip-plugin`	`xmip`	13	2	demo

Using Plugin Recipes

Installed plugin recipes are available through the same API as core recipes:

recipe = woodpecker.recipe.get("c3s.atlas")
preview = woodpecker.recipe.apply(dataset, recipe, dry_run=True)

The xMIP plugin is a demo of an xMIP-style CMIP6 preprocessing recipe expressed as small Woodpecker fixes:

recipe = woodpecker.recipe.get("xmip.cmip6_preprocessing")

Use Fix Reference and Recipe Reference for the full registered list.

C3S Contribution Pattern

C3S adaptations live in dataset-family plugins and are exposed through recipe ids:

Recipe id	Plugin	Purpose
`c3s.cmip6_decadal`	`roocs-woodpecker-cmip6-decadal-plugin`	CMIP6-decadal C3S/CDS preparation and adaptation.
`c3s.atlas`	`roocs-woodpecker-atlas-plugin`	Atlas C3S/CDS adaptation.

Contributors normally add one small adaptation class, register it in the plugin, add it to the recipe, and cover it with a synthetic dataset test. The recipe phase controls when it runs:

Phase	Use
`prepare`	Pre-concatenation changes.
`apply`	Normal adaptation steps. This is the default when `phase` is omitted.
`finalize`	Post-processing after normal adaptation.

For the CMIP6-decadal plugin, follow the module naming pattern documented in plugins/woodpecker-cmip6-decadal-plugin/README.md:

cmip6d_<sequence>_<short_name>.py

Labels

Labels are user-facing metadata for fixes.

Field	Purpose
id	Stable label identifier.
title	Short user-facing name.
description	Longer explanation.
category	Group such as `info`, `risk-low`, `risk-medium`, or `risk-high`.

Labels help users understand fixes. They do not affect recipes, priority, matching, or automation.

Plugins can use predefined labels:

from woodpecker.fixes import FixFunction, Labels


class RenameTempVariable(FixFunction):
    labels = [Labels.RISK_REVERSIBLE_RENAME]

Plugins can also register custom labels:

from woodpecker.fixes import LabelCategories, register_label

register_label(
    "my_plugin.experimental",
    "experimental",
    description="Early plugin fix that should be reviewed carefully.",
    category=LabelCategories.RISK_HIGH,
)