airflow

# Airflow Operations

**[AIRFLOW SKILL ACTIVE]** - Mention "Using Airflow MCP tools..." in responses.

## Tool Usage Rules

**Use Airflow MCP tools for all operations.** Never use:
- `astro dev run` commands
- `airflow` CLI commands
- Bash to read logs or check directories

MCP tools provide structured, reliable API access.

---

## Request Routing

Determine what the user wants and route accordingly:

### Simple Requests → Handle Directly with MCP Tools

For straightforward operations, call MCP tools directly using the routing table below.

### Complex Workflows → Delegate to Specialized Skills

For multi-step procedures, delegate to specialized skills:
- **Testing/Running DAGs**: `/data:testing-dags`
- **Debugging Failures**: `/data:debugging-dags`
- **Creating/Editing DAGs**: `/data:authoring-dags`

---

## MCP Tool Routing Table

Use this table to map user requests to the correct MCP tool:

| User Intent | Trigger Words | MCP Tool to Call | Notes |
|-------------|---------------|------------------|-------|
| **List all DAGs** | list, show, what dags, get dags, all dags | `list_dags` | Returns all DAGs with metadata |
| **Get DAG details** | show dag X, details for dag X, info about dag X | `get_dag_details(dag_id)` | Single DAG metadata |
| **Explore DAG** | what does dag X do, how does dag X work, show me dag X | `explore_dag(dag_id)` | DAG + tasks + source |
| **Get DAG source** | show code for dag X, source of dag X | `get_dag_source(dag_id)` | Python source code |
| **Test/Run DAG** | test dag, run dag, trigger dag, execute dag | `trigger_dag_and_wait(dag_id)` | Or delegate to `/data:testing-dags` |
| **Check DAG run status** | status of run X, how did run X go | `get_dag_run(dag_id, dag_run_id)` | Specific run details |
| **Debug failure** | why did dag fail, what went wrong, debug dag | `diagnose_dag_run(dag_id, dag_run_id)` | Or delegate to `/data:debugging-dags` |
| **Get task logs** | show logs for task X, task output, task errors | `get_task_logs(dag_i