fix(chart,engine): rename PROTECTOR_FALCO_ADDR -> PROTECTOR_BEHAVIOR_ADDR, drop compat fallback

README.md

@@ -120,7 +120,7 @@ like Cilium/Calico).
 | `PROTECTOR_ENGINE_ENABLE` | — | comma list of auto-applied action classes (`network`,`rbac`,`mount`,`identity`); empty = propose-only. Only `network` is live-actuatable; `escape` is never enableable. Add `judgement` to let the **model decide** a proven foothold (internet-exposed + KEV/critical CVE, e.g. log4shell): a cut requires the model's affirmative `exploitable` verdict — CVE *presence* alone is propose-only ([ADR-0013](docs/adr/0013-proof-winnows-model-decides.md); needs `network` to cut) |
 | `PROTECTOR_ENGINE_ACTUATOR` | `dryrun` | live-cut mechanism: `networkpolicy` (flannel/kube-router, e.g. k3s/k3d), `adminnetworkpolicy` (Cilium/Calico), `dryrun`. Unknown/empty fails safe to dry-run |
 | `PROTECTOR_ENGINE_JOURNAL_PATH` | — | decision-journal file on a mounted volume (PVC/hostPath). Appends each pass's breach verdicts + ledger apply/revert deltas (with revert reason) as JSON lines, size-rotated; replayed on boot so the findings snapshot, the judgement record, and the reversion log populate immediately after a restart. Unset/unwritable = in-memory only, no crash ([ADR-0015](docs/adr/0015-advisory-evidence-egress.md) mounted-volume posture) |
-| `PROTECTOR_BEHAVIOR_ADDR` | — | runtime-evidence ingest addr for the tool-agnostic behavioral port (the first-party agent, or any sensor, POSTs behaviors to `/behavior`); unset = no runtime feed. The deprecated `PROTECTOR_FALCO_ADDR` is still read as a fallback |
+| `PROTECTOR_BEHAVIOR_ADDR` | — | runtime-evidence ingest addr for the tool-agnostic behavioral port (the first-party agent, or any sensor, POSTs behaviors to `/behavior`); unset = no runtime feed |
 | `PROTECTOR_KEV_FILE` | — | CISA KEV catalogue path (JSON or newline CVE list); unset = no exploit intel |
 | `PROTECTOR_ENGINE_MODEL` | — | OpenAI-compatible chat-completions endpoint for the adjudicator (e.g. a local Ollama); unset = deterministic only, no adjudication |
 | `PROTECTOR_ENGINE_MODEL_NAME` | `qwen2.5:3b` | model name for the above |

charts/protector/README.md

@@ -144,7 +144,7 @@ be created in the enforced scope until it recovers. Bake a scope in `mode: audit
 
 ### Ingest authentication (on by default) — rollout ordering
 
-The engine's runtime/behavioral ingest (the `:9999` falco-ingest port) accepts
+The engine's runtime/behavioral ingest (the `:9999` behavior-ingest port) accepts
 observations that can make a proven attack chain *actionable*. App-layer
 authentication is **on by default** (`ingestAuth.enabled: true`): the chart
 provisions a Secret with a random bearer token, the engine **requires** it, and the

charts/protector/templates/agent-daemonset.yaml

@@ -1,7 +1,7 @@
 {{- if .Values.agent.enabled }}
 # protector's first-party eBPF behavioral collector (ADR-0014), a DaemonSet on every
 # node. It loads eBPF probes, resolves cgroup->pod, and POSTs normalized observations to
-# the engine's behavioral ingest (the falco-ingest service, /behavior). Passive and
+# the engine's behavioral ingest (the behavior-ingest service, /behavior). Passive and
 # read-only — it observes, it never blocks/kills/rewrites; enforcement stays the engine's
 # reversible NetworkPolicy cut.
 #
@@ -96,7 +96,7 @@ spec:
                   fieldPath: spec.nodeName
             # The engine's behavioral ingest base — the agent appends /behavior.
             - name: PROTECTOR_AGENT_ENDPOINT
-              value: "http://{{ include "protector.fullname" . }}-falco-ingest.{{ .Release.Namespace }}.svc.cluster.local:{{ .Values.engine.falco.port }}"
+              value: "http://{{ include "protector.fullname" . }}-behavior-ingest.{{ .Release.Namespace }}.svc.cluster.local:{{ .Values.engine.ingest.port }}"
             {{- if .Values.ingestAuth.enabled }}
             # Ingest authn (Fix A): present this bearer token on every POST so the engine
             # accepts the agent's observations. Same Secret the engine reads.

charts/protector/templates/deployment.yaml

@@ -198,9 +198,9 @@ spec:
             - name: https
               containerPort: {{ .Values.service.port }}
               protocol: TCP
-            {{- if and .Values.engine.enabled .Values.engine.falco.enabled }}
-            - name: falco-ingest
-              containerPort: {{ .Values.engine.falco.port }}
+            {{- if and .Values.engine.enabled .Values.engine.ingest.enabled }}
+            - name: behavior-ingest
+              containerPort: {{ .Values.engine.ingest.port }}
               protocol: TCP
             {{- end }}
             {{- if and .Values.engine.enabled .Values.engine.dashboard.enabled }}
@@ -310,9 +310,9 @@ spec:
             # when it does actuate.
             - name: PROTECTOR_ENGINE_ACTUATOR
               value: {{ .Values.engine.actuator | quote }}
-            {{- if .Values.engine.falco.enabled }}
-            - name: PROTECTOR_FALCO_ADDR
-              value: "0.0.0.0:{{ .Values.engine.falco.port }}"
+            {{- if .Values.engine.ingest.enabled }}
+            - name: PROTECTOR_BEHAVIOR_ADDR
+              value: "0.0.0.0:{{ .Values.engine.ingest.port }}"
             {{- end }}
             {{- if .Values.engine.dashboard.enabled }}
             # The read-only operator dashboard (ADR-0019). Zero-egress, in-cluster only —

charts/protector/templates/engine-services.yaml

@@ -1,21 +1,22 @@
-{{- if and .Values.engine.enabled .Values.engine.falco.enabled }}
-# Runtime-evidence ingest. A Falco/falcosidekick deployment (and the optional eBPF
-# agent) POSTs alerts/observations here. ClusterIP only — keep it reachable only from
-# in-cluster senders (a NetworkPolicy / mesh authorization is the floor).
+{{- if and .Values.engine.enabled .Values.engine.ingest.enabled }}
+# Runtime-evidence ingest. The first-party eBPF agent DaemonSet POSTs behavioral
+# observations here (ADR-0014's addendum: the Falco adapter that historically fed this
+# port has been retired). ClusterIP only — keep it reachable only from in-cluster
+# senders (a NetworkPolicy / mesh authorization is the floor).
 apiVersion: v1
 kind: Service
 metadata:
-  name: {{ include "protector.fullname" . }}-falco-ingest
+  name: {{ include "protector.fullname" . }}-behavior-ingest
   labels:
     {{- include "protector.labels" . | nindent 4 }}
 spec:
   type: ClusterIP
   selector:
     {{- include "protector.selectorLabels" . | nindent 4 }}
   ports:
-    - name: falco-ingest
-      port: {{ .Values.engine.falco.port }}
-      targetPort: falco-ingest
+    - name: behavior-ingest
+      port: {{ .Values.engine.ingest.port }}
+      targetPort: behavior-ingest
       protocol: TCP
 {{- end }}
 {{- if and .Values.engine.enabled .Values.engine.dashboard.enabled }}

charts/protector/values.yaml

@@ -273,7 +273,7 @@ webhook:
   enforcedFailurePolicy: Fail
 
 # --- ingest authentication (security fix) ---
-# The engine's runtime/behavioral ingest (the :9999 falco-ingest port) accepts
+# The engine's runtime/behavioral ingest (the :9999 behavior-ingest port) accepts
 # observations that can make a proven attack chain ACTIONABLE (the corroborated-now
 # signal). Any caller that can reach the port could otherwise forge those. This wires a
 # shared bearer token that the engine requires and the agent presents — app-layer
@@ -333,9 +333,11 @@ engine:
   #   dryrun             — log only; touches nothing (force shadow even under enforce).
   actuator: networkpolicy
   # Runtime-evidence ingest (the corroborated-now signal that makes a chain
-  # live-actionable). A Falco/falcosidekick deployment POSTs alerts here; the optional
-  # eBPF agent below also feeds this port. Enabling the listener is harmless on its own.
-  falco:
+  # live-actionable). The first-party eBPF agent below POSTs behavioral observations
+  # here (ADR-0014's addendum: the Falco adapter that historically fed this port has
+  # been retired — the agent is the sole deployed corroboration source). Enabling the
+  # listener is harmless on its own.
+  ingest:
     enabled: true
     port: 9999
   # Read-only operator dashboard (ADR-0019). A server-rendered, zero-egress view of the

docs/adr/0014-behavioral-telemetry-ebpf.md

@@ -281,3 +281,17 @@ when the Rust toolchain can emit BTF field relocations, the preflight retires wi
 offsets. Any comments claiming the object is "CO-RE-relocated against node BTF at load" are
 corrected to reflect this — baked offsets + a load-time BTF check, not relocation. See
 `docs/ideas/ebpf-offset-self-verification.md`.
+
+## Addendum — in-repo chart migrated to `PROTECTOR_BEHAVIOR_ADDR`; compat fallback dropped (2026-08-08)
+
+The rename addendum decision 1 above shipped `PROTECTOR_BEHAVIOR_ADDR` with
+`PROTECTOR_FALCO_ADDR` read as a deprecated compat fallback "until the deployed chart
+migrates." The **in-repo** `charts/protector` chart now renders only
+`PROTECTOR_BEHAVIOR_ADDR` (the `engine.falco.*` values key and the `…-falco-ingest`
+Service are renamed `engine.ingest.*` / `…-behavior-ingest` to match), and the engine
+(`engine/src/main.rs`) reads only `PROTECTOR_BEHAVIOR_ADDR` — the fallback is deleted.
+
+This addendum covers the in-repo chart only. The deployed cluster runs a diverged fork
+of this chart (outside this repository) that still sets `PROTECTOR_FALCO_ADDR`; porting
+the rename there, and rolling it out *before* the fallback-dropped engine, is a separate,
+deploy-ordering-sensitive step tracked outside this repo.

engine/src/main.rs

@@ -423,10 +423,9 @@ async fn run() -> Result<()> {
         let (active, scope) = posture.engine_arming();
         // Runtime-evidence ingest endpoint (the first-party agent, and any sensor, POSTs
         // behaviors here) for the RuntimeEvidence "corroborated-now" signal. Unset = no runtime
-        // feed. Prefer PROTECTOR_BEHAVIOR_ADDR; fall back to the deprecated PROTECTOR_FALCO_ADDR.
-        // compat: cluster chart still sets PROTECTOR_FALCO_ADDR; remove after the chart migrates.
+        // feed. The deprecated PROTECTOR_FALCO_ADDR compat fallback was dropped once the
+        // in-repo chart migrated (ADR-0014's addenda).
         let behavior_addr = env::var("PROTECTOR_BEHAVIOR_ADDR")
-            .or_else(|_| env::var("PROTECTOR_FALCO_ADDR"))
             .ok()
             .and_then(|v| v.parse::<SocketAddr>().ok());
         // The k8s audit-log ingest endpoint: the apiserver's audit webhook POSTs