net-kingdom/sso-mfa/k8s/bounded-executor/T06-PILOT-RUNBOOK.md

T06 runbook: the live pilot rotation under one admission (founder-run)

NK-WP-0056-T06. The executor rotates svc-authelia's LLDAP bind. Your only part in the run is the admission: you approve it at aal2 in informed-decision. Everything before that (sections 1 to 5) is one-time go-live of the standing paths. Section 7 is the run.

Each step prints a value-free result. Paste it back, and claude-code verifies it read-only and records it.

Commands:

• Admin steps use K="kubectl --kubeconfig $HOME/.kube/config-railiance01".
• OpenBao steps need bao login -method=oidc -path=netkingdom role=platform-admin (MFA) first.
• Always run git pull --ff-only in a repo before using it.

#	What	Where	Changes
1	Owner ingress (T16)	approval-engine, audit-core, rapp	two NetworkPolicies; read-only netcheck
2	Custody	railiance-platform	approve CCR-2026-0060, -0061, -0063; apply policies and roles, ESO; mint the KeyCape client secret
3	Accept the case	State Hub decision	security-cases/nk-bind-executor-approval.yaml becomes accepted
4	KeyCape client	net-kingdom	bind the env var and register the client in one KeyCape restart
5	Audit sender	audit-core	mint and seed the nk-bounded-executor token
6	Admission	net-kingdom (claude-code)	admissions/BEA-2026-0002.yaml for the window you choose
7	The run	rapp, net-kingdom	request, approve at aal2, open the window, run, close the window, verify

1. Owner ingress (T16)

Apply each owner's policy object on its own. Live approval-engine also has informed-decision-review-ingress, which is not in its file. Applying the whole file would therefore be a broader change than this step.

K="kubectl --kubeconfig $HOME/.kube/config-railiance01"
cd ~/approval-engine && git pull --ff-only
python3 -c "import yaml,sys; yaml.safe_dump([d for d in yaml.safe_load_all(open('deploy/networkpolicies.yaml')) if d and d['metadata']['name']=='approval-engine-bounded-executor'][0], sys.stdout)" > /tmp/ae-bx.yaml
$K apply --dry-run=server -f /tmp/ae-bx.yaml && $K apply -f /tmp/ae-bx.yaml

audit-core: run step 1 of audit-core docs/bounded-executor-receipts.md, "Founder runbook — nk-bounded-executor sender". It applies audit-core-nk-bounded-executor-ingress the same way.

Then prove the path from an executor-labelled pod. It holds no token and makes no login:

cd ~/rapp-bounded-executor && git pull --ff-only
make netcheck KUBECTL="$K"

Expected, on the last lines: "approval_engine_health": 200, "audit_core_health": 200 and "informed_decision_tcp": "ok", alongside the existing OpenBao, KeyCape and Authelia columns.

2. Custody (railiance-platform)

Follow railiance-platform's go-live runbook for the three requests (path in its hub message "NK-WP-0056-T06 custody go-live"):

• approve CCR-2026-0060, -0061 and -0063 with the approve tool;
• apply the policies, roles and ESO objects;
• mint the KeyCape client secret into platform/workloads/keycape/nk-bind-executor-approval without printing it;
• check that the ExternalSecret sso/keycape-nk-bind-executor-approval-client reports Ready=True.

Section 4 needs that Ready immediately before it runs.

3. Accept the service-client case

Answer State Hub decision 59d570ef ("Accept SecurityCase nk-bind-executor-approval"). claude-code then records the acceptance in the case file. The case type requires owner and founder acceptance; you are both.

4. KeyCape: bind and register in one restart

KeyCape runs one replica with strategy Recreate. Changing its env is itself a restart, and registering a confidential client whose secret is not bound makes the next start fail closed. So: pause the rollout, bind, register, then resume. That is one restart, with both changes (key-cape R4). Run it in a quiet minute: in-flight sign-ins are dropped.

cd ~/net-kingdom && git pull --ff-only
K="kubectl --kubeconfig $HOME/.kube/config-railiance01"
$K -n sso get externalsecret keycape-nk-bind-executor-approval-client -o jsonpath='{.status.conditions[?(@.type=="Ready")].status}'; echo   # True

C=$(mktemp); python3 sso-mfa/k8s/keycape/keycape-client.py client-from-case security-cases/nk-bind-executor-approval.yaml > "$C"
B=$HOME/.local/share/key-cape/verified-bin/e2b5fa1     # check-config binary from the deployed image (T09 runbook)
test -x "$B/keycape" || { mkdir -p "$B"; cid=$(docker create forgejo.coulomb.social/coulomb/key-cape@sha256:9e082dca78f65a4af3e92d3de3705b99e49ca9b298d9562ad0d37511c4306f8d); docker cp "$cid:/keycape" "$B/keycape"; docker rm "$cid"; }

# keycape-client.py uses its built-in kubeconfig through 127.0.0.1:16444, exactly as in the T09 runbook;
# it pins the cluster identity and refuses another cluster.
$K -n sso rollout pause deployment/keycape
$K -n sso patch deployment keycape --type=json -p '[
  {"op":"test","path":"/spec/template/spec/containers/0/name","value":"keycape"},
  {"op":"add","path":"/spec/template/spec/containers/0/env/-","value":{"name":"KEYCAPE_NK_BIND_EXECUTOR_APPROVAL_CLIENT_SECRET","valueFrom":{"secretKeyRef":{"name":"keycape-nk-bind-executor-approval-client","key":"client-secret"}}}}]'
python3 sso-mfa/k8s/keycape/keycape-client.py inspect --client-file "$C"
#   -> "deployment_env_bound": true, "change_needed": true; note before.uid and before.resource_version

Then put the noted values into the apply line, in place of UID and RV:

python3 sso-mfa/k8s/keycape/keycape-client.py apply --client-file "$C" \
  --case-acceptance "security-cases/nk-bind-executor-approval.yaml (accepted, hub decision in section 3)" \
  --check-config "$B/keycape" --expected-uid UID --expected-resource-version RV
$K -n sso rollout resume deployment/keycape
$K -n sso rollout status deployment/keycape --timeout=180s

If apply refuses, nothing in the registry changed. Undo the env patch before resuming: run $K -n sso rollout undo deployment/keycape, then $K -n sso rollout resume deployment/keycape, and paste the refusal.

claude-code then runs the type's verification. The executor's request Job (section 7) performs the token exchange per scope. The excess-scope refusal, the custody denials and the RBAC can-i denial are read-only checks.

5. Audit sender

Run steps 2 to 4 of audit-core's founder runbook (see section 1): preflight, the silent seed under warden access ... --exec, and its receipt. The seed needs CCR-2026-0063 applied (section 2).

6. Admission (claude-code prepares, you check)

claude-code runs, for the window you name (at most 2 h):

python3 bounded-executor/prepare_admission.py --id BEA-2026-0002 --commit 9bd7bf0 \
  --not-before 2026-11-05T09:00:00Z --hours 2 --out admissions/BEA-2026-0002.yaml

It commits the file. The pins are the files inside the pinned image (sha256:2926d7d9…, built from 9bd7bf0). Read it: the plan, the checks, the revert steps, the budget and the window are what you will approve.

7. The run

K="kubectl --kubeconfig $HOME/.kube/config-railiance01"
cd ~/rapp-bounded-executor && git pull --ff-only && cd ~/net-kingdom && git pull --ff-only
A=~/net-kingdom/admissions/BEA-2026-0002.yaml
CFG=$(mktemp --suffix=.json)
python3 -c "import json; c=json.load(open('$HOME/rapp-bounded-executor/examples/config.pilot.json')); c['trust_satisfied']=['runtime_secret_trust','runtime_identity_trust']; json.dump(c, open('$CFG','w'), indent=2)"
#   trust_satisfied is your attestation, recorded in the run's result: OpenBao custody and KeyCape identity are healthy

7a. Request. The executor files the approval as service:nk-bind-executor and presents it in informed-decision:

cd ~/rapp-bounded-executor && make request ADMISSION=$A CONFIG=$CFG KUBECTL="$K"
#   -> {"outcome": "presented", "approval_id": "...", ...}; note the approval_id

7b. Approve. Open the memo in informed-decision and approve it. You must be signed in at aal2 (OTP in this login).

7c. Open the window. This creates CCR-2026-0058's write role for this admission only:

cd ~/net-kingdom && bao login -method=oidc -path=netkingdom role=platform-admin
W=$(sha256sum ~/railiance-platform/credential-change-requests/CCR-2026-0058-*.yaml | cut -d' ' -f1)
J=$(sha256sum ~/railiance-platform/credential-change-requests/CCR-2026-0059-*.yaml | cut -d' ' -f1)
E=docs/evidence/$(date +%F)-nk56-t06-pilot; mkdir -p "$E"
python3 tools/service-binds/executor-lanes.py window-open --admission $A --write-ccr-sha256 "$W" \
  --journey-ccr-sha256 "$J" --receipt "$E/01-window-open.json"            # {"status": "window_opened", ...}

7d. Run. Put the approval id from 7a in place of APPROVAL:

cd ~/rapp-bounded-executor && make run ADMISSION=$A APPROVAL_ID=APPROVAL CONFIG=$CFG KUBECTL="$K"
make run-status ADMISSION_ID=BEA-2026-0002 KUBECTL="$K"
$K -n sso logs job/bounded-executor-bea-2026-0002 > ~/net-kingdom/$E/02-result.json

The last line of the log is the value-free result:

Exit	Result	Meaning
0	completed	rotated, delivered, restarted, journey passed
3	reverted	a check failed, and the previous value is back, checked
4	refused	nothing was spent
5	stopped, unknown or revert_failed	the alert names the step
6	consumed but not in custody	—

Kill switch, at any moment of the run:

$K -n sso delete job bounded-executor-bea-2026-0002

The executor finishes the step in flight, then reverts what it changed (grace period = the budget). Before consume it refuses and spends nothing.

7e. Close the window right after the result, or at window end at the latest:

cd ~/net-kingdom && python3 tools/service-binds/executor-lanes.py window-close --admission $A \
  --receipt "$E/03-window-close.json"                                      # {"status": "window_closed", ...}

7f. Tell claude-code. It does the rest read-only:

• it confirms the run's started event and receipt in audit-core (operator read by correlation_id);
• it validates the pair against receipt schema v0.1;
• it runs the posture-feedback report over the receipts (T07);
• it records the evidence and closes T06.