Pipelines
ReferencePipeline anatomyThe shape of a pipeline document: what it must declare, what it may declare, and what every step inherits from it.ReferenceScript stepsThe simplest step mode: a shell script in a container, optionally with its own image and mounted storage.ReferenceDependencies and parallelismHow the graph is ordered, how a step splits into concurrent tasks, and where the structural limits are.ReferenceStep outputsHow one step publishes a value and another consumes it, and the rules that make a reference valid.ReferenceVariables and scopesWhat a pipeline declares it needs, where those values resolve from, and how a scope decides the answer.ReferenceSecrets in stepsDeclaring the credentials a step needs, and what the platform does with them.ReferenceConditions and failure handlingSkipping work that does not apply, and deciding which failures should stop a run.ReferenceDisplay optionsChoosing how a run presents itself, in the UI and in the Git check-run summary.ReferenceApprovalsA durable human checkpoint in the middle of a run, and the rules that decide who may release it.ReferenceAI stepsReplacing a script with a goal, and choosing the model that pursues it.ReferenceAgent roles, knowledge, and toolsThe four separate controls over LLM behaviour: who it acts as, what it knows, what tools it may reach, and what may block it.ReferenceReusable steps and child pipelinesPulling in a shared step definition, and launching another pipeline as part of this one.ReferenceFinal deliverablesWhat the run produces for people to read after the work is done, including dashboard publications.ReferenceThe complete manifestEvery capability in this chapter assembled into one pipeline, with the reference tables that carry the rest.