Apache Airflow package v1.0.0

List the DAGs in the deployment's metadata database; this pack registers the collection read only — single-DAG reads (GET /api/v1/dags/{dag_id}) and GET /api/v1/dags/{dag_id}/details are not registered. DAG names, tags and schedules reveal the organization's pipeline layout.

The host is a reserved placeholder: substitute the customer's webserver host before enabling.

List DAG runs for one DAG; this pack registers the collection read only — single-run reads (GET /api/v1/dags/{dag_id}/dagRuns/{dag_run_id}) are not registered. The API accepts '~' as the dag_id to list runs across all DAGs; whether the deployment permits that pattern is a binding decision, not a restriction this registration imposes. DAG_ID is a placeholder the deployment binds before enabling.

Batch-list DAG runs across many DAGs. This is a POST-with-query-body endpoint used read-only: the specification defines it as a POST only because a GET filtering across a large number of DAG ids would exceed HTTP request URL length limits, and the request body is a query form (ListDagRunsForm: dag_ids, states, execution-date range filters, page limit and offset). It changes nothing and is registered read at the minimum R0 tier.

The host is a reserved placeholder: substitute the customer's webserver host before enabling.

List the stored Airflow connections; this pack registers the collection read only — single-connection reads (GET /api/v1/connections/{connection_id}) are not registered. Connection records describe the infrastructure the deployment reaches (types, hosts, ports, logins and extras), which is sensitive internal configuration even where secret fields are not rendered; treat the response accordingly.

The host is a reserved placeholder: substitute the customer's webserver host before enabling.

Update one DAG record — the operational use is pausing and unpausing via is_paused with the update_mask query parameter. Unpausing is not an ordinary reversible update: it releases the DAG's queued and scheduled execution, and the tasks then run against whatever systems the DAG is wired to, so the effect is deploy-equivalent and the capability declares update and deploy at the minimum R3 tier. This registration does not restrict which fields the request body may carry: an installer's intention of pause-only is not a restriction, and a reviewer who wants that must impose it in policy.

The collection-wide PATCH /api/v1/dags (dag_id_pattern, including '~' for all DAGs) is a different, bulk operation and is deliberately not registered by this pack. DAG_ID is a placeholder the deployment binds before enabling.

On supported execution paths mediated by AAES Govern, the declared irreversible actions require an authorized person’s approval. Observation-only and bypass calls cannot be stopped by this control.

Create one stored connection in the Airflow metadata database.

The request body is a Connection record and can carry credential material (login, password, extra) that Airflow then stores and serves to tasks — the body of a call to this capability is itself sensitive: the executor must keep the request body out of logs, traces, diagnostics and crash dumps exactly like the injected Authorization material, must not buffer it beyond the lifetime of the call, and any deployment-side request recording must redact it.

Creation is reversible by deletion, but any secret written is thereafter held in Airflow's own store, outside this capability's reach; review what the integration is permitted to write before enabling.

The host is a reserved placeholder: substitute the customer's webserver host before enabling.

Trigger a new DAG run. This is registered with the deploy action at the minimum R3 tier rather than as an ordinary create: a DAG run executes the DAG's tasks, which can write to any system the DAG is wired to — production databases, warehouses, third-party APIs — so the effect is deploy-equivalent, and the request body's conf object is an unrestricted parameter payload handed to that execution.

Deploy is one of the designated actions requiring human approval.

On supported execution paths mediated by AAES Govern, irreversible actions require approval by an authorized person; this does not prevent calls through bypass or observation-only paths. If the DAG is paused the run remains queued per the API's documented behavior, which changes timing, not effect. DAG_ID is a placeholder the deployment binds before enabling.

Delete a DAG's metadata. The OpenAPI description is explicit: this deletes all metadata related to the DAG, including finished DAG runs and tasks, logs are not deleted, and the action cannot be undone — the run history is the audit trail for past executions, and re-adding the DAG code later does not restore it. Registered as delete at the minimum R3 tier; delete is one of the designated actions requiring human approval. DAG_ID is a placeholder the deployment binds before enabling.

On supported execution paths mediated by AAES Govern, the declared irreversible actions require an authorized person’s approval. Observation-only and bypass calls cannot be stopped by this control.

Delete a stored connection, including any credential material held in it. Recreating the connection later is a new write, not an undo: unless the caller preserved the previous secret values elsewhere they are gone, and every DAG that references the connection id starts failing at runtime — the blast radius extends beyond the deleted record. Registered as delete at the minimum R3 tier; delete is one of the designated actions requiring human approval. CONNECTION_ID is a placeholder the deployment binds before enabling.

On supported execution paths mediated by AAES Govern, the declared irreversible actions require an authorized person’s approval. Observation-only and bypass calls cannot be stopped by this control.