azure-pipelines-node-ci

Azure Pipelines for a Node.js project

For every task, apply Caveman. Before broad repository navigation, use CocoIndex Code; for web-page research, use Crawl4AI. Read and follow the repository-root docs/agent-tools.md and the three linked skills, and retain the required execution evidence.

This maintained overlay covers a basic lint-and-test pipeline for a project with an npm lockfile, a lint script, a test script that is exactly node --test, and Python 3 available as python3 on the agent. It does not cover deployment, other package managers, other test frameworks, coverage reporting or Azure DevOps Server. Inspect the project's actual scripts, runtime requirements and supported Azure environment before using the example; do not assume that arbitrary npm test scripts accept the same flags.

Start from the example only when its prerequisites match. It pins Node 26.13.0 and the Microsoft-hosted ubuntu-24.04 image; choose a maintained version compatible with the project and validate changes. Install from the existing npm lockfile, fail the job on lint or test failures, and publish the native test runner's JUnit output. The example requires at least one non-skipped test case; that alone does not prove the complete intended suite ran. A missing result file or a failure to publish is an error. Add project-specific build and coverage steps only when the project provides and verifies their actual outputs.

Run the project's commands in an isolated copy, including deliberate lint and test failures, and inspect the resulting JUnit file. Validate and run the YAML in the intended Azure organization and repository before relying on its hosted behavior. The recorded controls exercise the local commands and parse the YAML; they did not execute an Azure pipeline. Azure Repos pull-request validation requires an appropriate branch policy; configure triggers and policies for the repository provider and user requirements.

The original upstream instructions and example are retained under provenance/. The overlay's component controls are not a coding-agent benchmark or proof that the full upstream skill works.