Project Configuration
Each repository can have a .azent.json file in its root directory to configure project-specific behavior.
File format
{
"ValidationPipelineId": 42,
"Preprompt": "This is a .NET 8 REST API project using minimal APIs..."
}
Fields
| Field | Type | Description |
|---|---|---|
| ValidationPipelineId | number | null | The Azure Pipelines pipeline ID to run for validation after changes. See Validation Pipeline. |
| Preprompt | string | null | Custom instructions appended to the agent prompt for this specific project. Use this for project-specific context the agent should always have. |
Preprompt examples
The preprompt is a great place to give the agent context it can't easily derive from the code alone:
{
"Preprompt": "This is a .NET 8 REST API. Use minimal APIs, not controllers. All endpoints return IResult. The database is accessed through repository classes in src/Data/. Always add unit tests in the Tests project for new endpoints."
}
Note: The preprompt is project-specific and complements the workspace-level skills. Skills apply to all projects in the workspace, while the preprompt applies only to the repository that contains the .azent.json file.
Agent instruction files
Each AI agent also reads its own conventional Markdown file from the repository root. These are picked up automatically — no .azent.json entry required — and are a natural place to keep coding standards, architectural notes, and contributor guidance versioned alongside the code:
| Agent | File |
|---|---|
| Codex (OpenAI) | AGENTS.md |
| Claude (Anthropic) | CLAUDE.md |
If both files exist, the agent reads the one that matches the active provider. Keep these files in sync if you switch providers regularly, or symlink one to the other.