Skip to content

Bitbucket Pipelines

Status: AI draft, 2026-11-16. Not yet verified by me. Written from the Bitbucket Cloud documentation; nothing was run.

In one line

Bitbucket Pipelines is the CI/CD built into Bitbucket Cloud: bitbucket-pipelines.yml defines pipelines per branch, pull request, tag or manual trigger, each step runs in a fresh container, and deployments, caches, artifacts, pipes and oidc: true cover environments, speed, reuse and keyless AWS access.

How it works

flowchart TD
    E["Push, pull request, tag,<br/>schedule or manual run"] --> P["bitbucket-pipelines.yml:<br/>pick the matching pipeline"]
    P --> S1["Step 1 in a fresh container:<br/>script, caches"]
    S1 -->|"artifacts"| S2["Step 2: deployment: staging"]
    S2 --> S3["Step 3: deployment: production,<br/>trigger: manual"]
image: node:22

definitions:
  caches:
    npm-cache: ~/.npm

pipelines:
  pull-requests:
    "**":
      - step:
          name: Test
          caches: [npm-cache]
          script:
            - npm ci
            - npm test
  branches:
    main:
      - step:
          name: Build
          caches: [npm-cache]
          script:
            - npm ci
            - npm run build
          artifacts:
            - dist/**
      - step:
          name: Deploy to staging
          deployment: staging
          oidc: true
          script:
            - ./deploy.sh staging
      - step:
          name: Deploy to production
          deployment: production
          trigger: manual
          oidc: true
          script:
            - ./deploy.sh production
Concept Meaning
image Default Docker image for steps (each step may set its own)
pipelines default, branches, pull-requests, tags, custom (manual or scheduled)
step One container run; script is a list of shell commands
parallel Steps that run at the same time
caches Predefined (such as node, pip) or custom directories kept between runs
artifacts Files passed from one step to later steps in the same pipeline
deployment Marks the step as a deploy to an environment (test, staging, production types), shown on the Deployments page and given that environment's variables
trigger: manual Wait for someone to click before running the step
Variables Workspace, repository and deployment level; secured variables are encrypted and masked in logs
pipe A packaged task in a container, such as atlassian/aws-s3-deploy, configured with variables
oidc: true The step gets a token in BITBUCKET_STEP_OIDC_TOKEN for AWS role assumption (OIDC from pipelines to AWS)
Bitbucket Pipelines GitHub Actions
File One bitbucket-pipelines.yml Many files in .github/workflows/
Unit of work Step = one container Job = one runner with steps
Reuse Pipes, YAML anchors, shared pipeline configurations Actions, reusable workflows (GitHub Actions in depth)
Environments Deployments with variables, permissions and manual triggers Environments with protection rules
Keyless AWS oidc: true id-token: write
Self-hosted Runners Self-hosted runners

Interview questions

1. Describe the structure of bitbucket-pipelines.yml.

Answer
  • An optional default image, definitions (caches, services, YAML anchors), and pipelines.
  • Pipelines by trigger: default, branches, pull-requests, tags, custom.
  • Each contains steps with a script, optional caches, artifacts, deployment and trigger.

2. What is the difference between caches and artifacts?

Answer
  • Caches keep dependency directories between pipeline runs to save download time; not guaranteed.
  • Artifacts pass build outputs from one step to later steps of the same run.
  • Never rely on a cache for correctness.

3. How do deployments and variables work together?

Answer
  • A step with deployment: production is tracked on the Deployments page and receives that environment's variables.
  • Secured variables are encrypted and masked.
  • Combined with trigger: manual and branch permissions, this gates production.

4. How do you deploy to AWS from Bitbucket without stored keys?

Answer
  • Register the workspace's OIDC provider in IAM and create a role trusting it.
  • Pin the trust policy's sub to the repository UUID and, for production, the deployment environment UUID.
  • Add oidc: true to the step and give the token to the AWS CLI or SDK through AWS_WEB_IDENTITY_TOKEN_FILE and AWS_ROLE_ARN.

5. What is a pipe?

Answer
  • A reusable, containerised task called with pipe: and configured with variables.
  • Examples: deploying to S3, notifying Slack, scanning.
  • Pin the pipe's version, as with any third-party action.

Sources