Find what broke the build, not just that it broke

**Objective:** Work out why `{pipeline-name}` in project `{project}` (all projects if empty) is failing and what changed when it started.

**Step 1: Resolve the pipeline**

Call `azure-devops-get-build-definition` with name.contains `{pipeline-name}` and `azure-devops-get-pipeline` with name.contains `{pipeline-name}`. If several match, list them and stop. Use the definition id for classic builds and the pipeline id for YAML pipelines.

**Step 2: Read recent runs**

For a build definition call `azure-devops-get-build` with definition-id; for a YAML pipeline call `azure-devops-get-pipeline-run` with pipeline-id. Take the latest 20 runs and find the first failing run after the last success, and the failure rate.

**Step 3: Find what changed**

Call `azure-devops-get-commit` for the repository-id of the first failing build (from `azure-devops-get-build`) with author-date.from set near that run. List commits between the last green run and the first red one with author and message.

**Step 4: Report**

Failure rate over the runs examined, the first failing run, the commits that landed between the last success and it, and a ranked list of likely causes with the evidence for each.

**Rules & Guidelines:**

* This prompt is read-only. Call only the `azure-devops-get-*` read tools named above; never create, update, comment on, approve, queue, abort or delete anything in Azure DevOps.
* The tools only return data the connected account can see; never infer data from projects outside that access.
* Say plainly that build logs are not available through these tools; do not guess at error messages.
* Distinguish a flaky pipeline (alternating results) from a broken one (consecutive failures).

**Validation:**

* If the pipeline has no runs, report that and stop.
* If every recent run succeeded, say it is healthy.