docs: aviation flavor — airspace regions, ops manual, Airworthiness Directives

README.md

@@ -83,6 +83,12 @@ One agent per **lane** (a domain of work). Each runs on an elastic cloud node,
 is scoped to its lane and nothing else, and actually writes and pushes code.
 Add capacity by adding lanes, not by cloning one agent.
 
+Each lane is its own **region of airspace**: it flies under its own flight
+rules — its scope, its conventions, what it's cleared to touch — and an executor
+is qualified for that region and no other. The procedures an agent follows — its
+**skills** — are its operations manual: the standard way to do the job, not
+improvised each run.
+
 ---
 
 ## How work flows

docs/agentic-iac.md

@@ -48,6 +48,12 @@ flavor, not the mechanism.
   Cat B = 3 days, C = 10, D = 120) before it **must** be fixed or the aircraft
   can't dispatch. Deferring infra work is exactly that: a logged snag with a
   clock on it, not a shrug.
+- Not everything can wait, though. A security update or a must-do change is an
+  [**Airworthiness Directive**](https://skybrary.aero/articles/airworthiness-directive) —
+  mandatory, from the authority, with a compliance deadline (an *emergency* AD
+  means *now*). Unlike an MEL snag you can't ride it out: an overdue AD grounds
+  the aircraft, and the release below **cannot even be signed** while one is
+  outstanding. A critical patch lands before anything else ships.
 - The infra agent **turns the wrench** — it does the maintenance and writes the
   CDK. What it cannot do is sign the aircraft back into service. A
   [**Certificate of Release to Service**](https://skybrary.aero/articles/certificate-release-service-crs)