the-custodian/docs/changes/netkingdom-target-2026-10-16/group-postgres-r1/final-integration-r1/workplan.postimage.md

---
id: NK-WP-0048
type: workplan
title: "Establish SecurityCaseTypes so that known security cases are light"
domain: infotech
repo: net-kingdom
status: active
flavor: implementation
owner: claude-code
topic_slug: netkingdom
created: "2026-10-08"
updated: "2026-10-17"
depends_on: []
related_repos:
  - ops-mason
  - ops-warden
  - key-cape
  - railiance-platform
  - informed-decision
state_hub_workstream_id: "e31f3f9d-0748-5137-b0a3-7b3ca46a66eb"
---

# NK-WP-0048 - Establish SecurityCaseTypes so that known security cases are light

## Why

The founder's direction (2026-10-08): setting up a security integration of a
kind that has been done before should be easy, meaning accept the case, then
ops-mason arranges it and ops-warden follows through. Setting up a new kind
needs ample documentation and tests and an informed-decision, because its shape
is part of the security architecture.

Today every integration is hand-built: per-client helper scripts, per-lane
custody, founder checklists. The trigger was KEY-WP-0038 (Forgejo as an OIDC
client), whose next step was going to be yet another hand-written rollout
script.

This plan continues the pattern work of NK-WP-0008 (pattern infospace),
NK-WP-0009 (tutorials) and NK-WP-0030 (scenario composition). It turns patterns
into **SecurityCaseTypes** that ops-mason and ops-warden can execute. Canon:
`canon/standards/security-case-types_v0.1.md` (proposed).

Companion plans: ops-mason `MASON-WP-0010` (materialization) and ops-warden
`WARDEN-WP-0044` (follow-through).

## Tasks

```task
id: NK-WP-0048-T01
status: done
priority: high
state_hub_task_id: "930ac3e9-20ec-5473-98cd-3fe8fddb730f"
```

### T01 — Accept the terminology and the canon

Founder review of `canon/standards/security-case-types_v0.1.md`, including its
section 10 open questions. On acceptance the canon status goes from `proposed`
to `accepted`.

2026-10-08: accepted by the founder (in session). Canon status `accepted`;
section 10's open questions are carried forward to the first type (T04, T05).

```task
id: NK-WP-0048-T02
status: done
priority: high
state_hub_task_id: "19c25d96-a55c-58f0-9168-3fdbc4955fc2"
```

### T02 — Inventory the de-facto types already built by hand

Read-only. List every security integration built so far and group them by shape.
Known starting points:

- browser OIDC relying parties on key-cape: informed-decision-approver,
  coulomb-social, user-engine portal, openbao-admin, and Forgejo (pending)
- confidential service clients with `client_credentials`: secrets-engine
  approval, requester and openbao clients, approval-engine operator
- workload KV read lanes on OpenBao: the railiance-platform credential lane
  designs and CCRs, for example `railiance-backup-offsite-lane`
- short-lived SSH certificates through ops-warden (`warden sign`)
- tunnelled operator access (ops-bridge)

For each: the repositories involved, the hand-built steps, attended human steps,
custody paths, verification evidence, and rollback (if any). Output
`docs/security-case-type-inventory.md`, ranking candidate types by how often the
shape recurs and how much manual work each instance costs today.

2026-10-08: done, `docs/security-case-type-inventory.md`. 36 of 37 CCRs are
already `request_type: workload-kv-read`, a de-facto type reviewed each time as
if it were new. Browser clients: 5 hand-built plus Forgejo, each with its own
rollout script. Confidential service clients need a registration plus two or
three KV-read CCRs each. A forge read identity is needed in bulk by
ORGREF-WP-0002-T06. Recommendation: oidc-browser-relying-party first (unblocks
the forge), workload-kv-read second (largest payoff), then forge-read-identity.

```task
id: NK-WP-0048-T03
status: done
depends_on: [NK-WP-0048-T01]
priority: high
state_hub_task_id: "e487975b-73b1-516c-971d-da4cbbb93063"
```

### T03 — Schema, template and validator

- `canon/schemas/security-case-type_v0.1.schema.json` and
  `security-case_v0.1.schema.json`
- `security-case-types/_template/` with `type.yaml`, `README.md` and tests to copy
- a validator, extending `tools/security-scenario-composer/`, that checks a type,
  checks a case against its type, and renders the case into the composer's
  deterministic plan (still `execution.permitted: false`)

Acceptance: the validator rejects missing sections, secret-shaped parameter
values, unpinned providers and cases against non-`implemented` types, with tests
for each.

2026-10-08: done. `canon/schemas/security-case-type_v0.1.schema.json` and
`security-case_v0.1.schema.json`; `tools/security-case-types/security_case_validator.py`
(`validate-type`, `validate-case`, `render`, reusing the scenario composer);
`security-case-types/_template/` (a valid, runnable example type over
`net-kingdom.local-identity`, with README sections, an example case and tests that
travel with every copy). 31 tests pass, including the composer's existing suite.

```task
id: NK-WP-0048-T04
status: done
depends_on: [NK-WP-0048-T02, NK-WP-0048-T03]
priority: high
state_hub_task_id: "2fffe112-7d55-5680-9952-78d74f44f87f"
```

### T04 — Define the first type: browser OIDC relying party

`security-case-types/oidc-browser-relying-party/`, version 1.0:

- parameters: client id, display name, exact redirect URIs, scopes, client type
  (confidential or public), MFA pin, the groups the consuming service maps,
  custody path for the secret (confidential only), consuming service and its
  owner
- composition: key-cape client registration (key-cape declares it as a playbook
  capability), secret minting into custody (railiance-platform), LLDAP group
  existence and membership (net-kingdom identity-provisioner), handover to the
  consuming service
- materialization (ops-mason) and follow-through (ops-warden) sections
- verification: discovery lists the redirect URI; a test sign-in returns the
  expected claims for one human; anonymous access to the custody path is denied
- rollback: remove exactly this client and its custody path
- README: threat model, MFA rationale, why redirect URIs are exact, which steps
  still need an attended login and how that is minimised
- reference instances: informed-decision-approver, coulomb-social, user-engine
  portal, openbao-admin

Acceptance: complete per canon section 3, tests green, status `proposed`.

2026-10-08: done. `security-case-types/oidc-browser-relying-party/` 1.0 (proposed):
`type.yaml` with 12 parameters (the custody reference is required only for
confidential clients), composition over key-cape's new
`key-cape.oidc-client-registration` declaration (contract PASS, `maturity:
planned`), materialization and follow-through, verification and rollback, all
conditional on client type where it matters. One attended step,
`register-client`, justified by the live registry being a Kubernetes Secret.
README with threat model, alternatives, an honest fit table (portal and vergabe
fit; Forgejo is the first confidential case; openbao-admin, coulomb.social and
informed-decision-approver do not fit 1.0), and known limits. Examples: Forgejo
(checked against the accepted request) and a public portal. The validator gained
`when` / `required_when` conditions and provider discovery. 43 tests pass.
ops-mason's and ops-warden's sections are drafted in the type. Their owners
review them in MASON-WP-0010-T01 and WARDEN-WP-0044-T01 before acknowledgment
(T05).

```task
id: NK-WP-0048-T05
status: done
depends_on: [NK-WP-0048-T04, MASON-WP-0010-T01, WARDEN-WP-0044-T01]
priority: high
state_hub_task_id: "3d987d3d-0797-538b-86e4-51185a327ae2"
```

### T05 — Acknowledge the type through informed-decision

Prepare the Decision Memo from the type's README and `type.yaml` and submit it
through informed-decision's production intake. On acknowledgment the type
moves to `acknowledged`.

2026-10-08: Decision Memo prepared. Hub decision `e418263e-a71f-4f01-a970-c87c9e187c09`.
`security-case-types/oidc-browser-relying-party/decision/`: `packet.md` (question,
one act, what acknowledging does not do, the shape, alternatives, attended
steps, limits, evidence), `build_intake.py` and the built `intake.json` (memo
class `decision-record`, review group `net-kingdom-admins`, resolves the hub
decision on answer). The envelope passes informed-decision's own
`validate_submission`. Queuing it is an attended operator step, a write inside
the production dispatcher pod:

    kubectl -n state-hub exec -i deploy/state-hub-memo-dispatch -- \
      python -m api.services.memo_dispatch --queue /data/private enqueue --envelope /dev/stdin \
      < security-case-types/oidc-browser-relying-party/decision/intake.json

The answer then appears for review in informed-decision.

2026-10-08: queued by the founder (`pending`, `queued`) after one refusal: the
first envelope's idempotency key was not State Hub's stable
`state-hub.<decision>.<memo>` form, and the builder now validates against both
sides. The dispatcher delivered it to informed-decision (ledger 7 → 8
delivered). T05 waits for the founder's answer in informed-decision, which
resolves hub decision `e418263e`.

2026-10-08: ops-mason's review (MASON-WP-0010-T01) folded into the type, still
1.0 `proposed`: `secret_ref` is no longer a parameter, the type derives
`platform/workloads/keycape/{client_id}`; policies are
`workload-kv-read-keycape-{client_id}-eso` and
`workload-kv-read-keycape-{client_id}-{consuming_service}`, each with a templated
body (`read` on that one data path); new `openbao_role`
`external-secrets-keycape-{client_id}` (Kubernetes auth, ESO service account,
15m, no default policy); catalog entry `{consuming_service}-keycape-client`;
the two rollback steps on the KV path collapse into one metadata delete. The
schema and validator now require a `body` on policy and role items. 105 tests
pass. Two consequences: the delivered memo describes the type before these
changes (a case-set custody reference, two differently named policies, no ESO
role), so the founder answers on a superseded shape unless the memo is
re-sent; and Forgejo's custody path moves from the accepted request's proposed
`platform/workloads/forgejo/netkingdom-oidc-client` to
`platform/workloads/keycape/forgejo`, which key-cape (KEY-WP-0038) and
railiance-apps (RAPPS-WP-0016-T02) need to adopt in T06. ops-warden's review
(WARDEN-WP-0044-T01) is still in progress.

2026-10-08: revised memo `nk-wp-0048-t05-oidc-browser-relying-party-r2` built
(`packet.md`, `intake.json`, packet `sha256:af6435e2…`); it states that it
supersedes the first memo and passes informed-decision's and the dispatcher's
validation. It needs a new memo id because the dispatcher refuses new content
under a delivered idempotency key (`idempotency_conflict`). Queuing it is the
same attended step as above. Both memos serve hub decision `e418263e`; the
first answer resolves it, so the first memo must not stay open: an accept on it
would acknowledge the superseded shape, and a decline would resolve the
decision against r2. Close it with `return` (a reviewer verb that requests a
new revision and does not resolve the hub decision; only accept and decline
do), commenting that r2 supersedes it, then answer r2. `withdraw` exists in
the vocabulary but is not offered to reviewers, and lifecycle `revise` only
applies to lifecycle-registered memos, not intake-delivered ones.

2026-10-09: the founder reports r2 enqueued and the first memo returned. At
close, hub decision `e418263e` still showed `open` (not verified whether r2
was answered or the answer is still in transit; the dispatcher ledger was not
read). ops-warden's review (WARDEN-WP-0044-T01, ops-warden
`history/2026-10-08-oidc-rp-follow-through-review.md`, d470f2e) arrived at
21:25, before r2 was built, and is not in r2. Required before WARDEN-WP-0044-T02:
R1 a per-step `lane` (missing: identity-provisioner caller credential,
railiance-platform broker grant), R2 canon meaning of `warden_executes: true`
(caller proxy, caller identity, no standing ops-warden token; ADR-0002), R3 a
per-step `value_handling`. Refinements R4 (one type-level catalog lane instead
of a per-case entry, which conflicts with ops-mason's item 5), R5 (ESO Ready gate
before bind-env), R6 (exact accepted memberships), R7 (mfa-pinned and
secret-delivered checks). Next: if r2 was accepted, fold R1-R3 and R5-R7 in as
a follow-up and settle R4 between ops-warden and ops-mason before T06; if not,
hold r2, fold them in and build r3.
 key-cape and railiance-apps were told about Forgejo's new custody
path.

2026-10-09: r2 was not answered (hub decision `e418263e` open; dispatcher 9
delivered), so r2 is held and r3 replaces it. ops-warden withdrew R4 for R4′
(message a66a9204): one catalog entry per case with its body templated in the
type. Folded in:

- Canon section 3: a narrow meaning of `warden_executes: true` (caller proxy,
  caller identity, no standing credential; ops-warden ADR-0002), plus per-step
  `lane`, `value_handling` and an optional `ready_check`. A catalog entry body is
  now required, like policy and role bodies.
- Schema and validator enforce all of these.
- The type: lanes `identity-provisioner-groups` and
  `keycape-client-secret-generate` (both new) and `key-cape-oidc-login`;
  `mint-secret` is `generated_in_custody`; a Ready gate on `deliver-secret`;
  `ensure-groups` adds exactly the accepted case's memberships; new checks
  `secret-delivered` and `mfa-pinned`; the catalog-entry body follows R4′.
- 110 tests pass.

Memo r3 (`nk-wp-0048-t05-oidc-browser-relying-party-r3`, packet
`sha256:726dfc01…`) passes informed-decision's and the dispatcher's validation.
The founder queues it with the same command, returns r2 with a comment that r3
supersedes it, then answers r3.

Prerequisites for T06, not for T05:
- the two new lanes, each through its owner;
- a founder-approved rollout of an identity-provisioner image containing
  630ebfd (the deployed one is main-6c4fcaf from 2026-09-24, without
  `/v1/groups/ensure`).

ops-mason's request to create LLDAP group `fluid-telegram-operators` (message
259ec26d) was answered: not possible from here by that path; the founder creates
it in the LLDAP admin UI.

2026-10-09: ops-warden finished its side of WARDEN-WP-0044-T01 (message
`58d33f40`, commit 0e7ceb8): the revised type conforms to `warden case
check-type`, with no further changes asked of the type. This acknowledges that
review. Its later addendum R8 (message `af274783`, ops-warden
`history/2026-10-08-oidc-rp-follow-through-review.md`) showed that lane
`identity-provisioner-groups` cannot exist as declared: the provisioner's only
HTTP credential is user-engine's runtime token, and its NetworkPolicy admits
only user-engine pods, so a caller-proxy lane would act as user-engine. Taken
option 1 (ops-warden's recommendation): `ensure-groups` and its rollback are
attended, in the `register-client` session on `key-cape-oidc-login`, run inside
the provisioner pod; no token is read. Option 2 (an operator credential of its
own on the provisioner) stays open for a later version. `warden case
check-type` now reports only `keycape-client-secret-generate`
(railiance-platform) as missing. 89 tests pass.

Hub decision `e418263e` was still open, so memo r4
(`nk-wp-0048-t05-oidc-browser-relying-party-r4`, packet `sha256:452ea421…`)
replaces r3; it passes informed-decision's and the dispatcher's validation. The
founder queues it with the same command, returns r2/r3 if still open with a
comment that r4 supersedes them, then answers r4.

Prerequisites for T06 are now: lane `keycape-client-secret-generate`
(railiance-platform), and a founder-approved rollout of an identity-provisioner
image containing 630ebfd.

2026-10-09: done. The founder queued r4 (`pending`, `queued`), returned r3 as
superseded and accepted r4 in informed-decision. informed-decision's hub
write-back is off in production (`INFD_HUB_WRITEBACK` unset), so net-kingdom
resolved hub decision `e418263e` from the founder's report, naming memo r4 in
the rationale. The type is `acknowledged`; it becomes `implemented` with T06.

```task
id: NK-WP-0048-T06
status: wait
depends_on: [NK-WP-0048-T05, NK-WP-0048-T08, NK-WP-0048-T09, NK-WP-0048-T12, NK-WP-0048-T13, NK-WP-0048-T14, NK-WP-0048-T15, NK-WP-0048-T16, NK-WP-0048-T17, NK-WP-0048-T18, NK-WP-0048-T20, MASON-WP-0010-T03, WARDEN-WP-0044-T03]
blocking_reason: "Type1.2 architecture has a confirmed source-only founder acknowledgment; canonical1.1 remains unchanged/default and the1.2 reference case stays proposed. Needs explicit version selection/reference-source integration without normal admission bypass, closed owner adapters plus global/native receipt verifier (WARDEN-WP-0044-T03), independently verified six-object structural materialization and CCR42 claims/created-only provenance, exact broker/case/source reconciliation, then separately authorized native mint/delivery and attended registry/consumer verification. Source acknowledgment, a bundle or local tests do not prove implemented/materialized/live status."
priority: high
state_hub_task_id: "d9b26bcf-a7c6-59a2-92a0-f913d76a203c"
```

### T06 — First KnownSecurityCase: Forgejo

Write `security-cases/forgejo-netkingdom-oidc.yaml` from the parameters the
founder accepted in KEY-WP-0038-T01 (`key-cape/docs/forgejo-oidc-client-request.yaml`),
so that acceptance carries over as the case acceptance. ops-mason materializes it;
ops-warden follows through and verifies it. The type becomes `implemented` once
this reference case is live.

This replaces the hand-built path proposed for KEY-WP-0038-T02.

2026-10-09: case written, `security-cases/forgejo-netkingdom-oidc.yaml`, status
`accepted`: the example's parameters plus the founder's 2026-10-08 acceptance of
the key-cape request (owner and founder), with the custody-path derivation
covered by the type's acknowledgment (memo r4). It validates and renders (plan
`execution.permitted: false`; three attended steps on `key-cape-oidc-login`);
it needs `--allow-unimplemented` until it is live, since it is the case that
makes the type `implemented`. A test keeps it equal to the type's example.
Still open before it runs: ops-mason renders and builds it (MASON-WP-0010-T02,
T03, unblocked by T05), ops-warden plans and runs it (WARDEN-WP-0044-T02, T03),
railiance-platform's `keycape-client-secret-generate` lane (asked by ops-mason
`789feb39` and ops-warden `d9f889b1`, both unread), and a founder-approved
identity-provisioner rollout containing 630ebfd.

2026-10-09: identity-provisioner rolled out, founder-approved in session:
`main-630ebfd` (`sha256:bafb4860…`, built by the image workflow) replaces
`main-6c4fcaf` (`sha256:ffacd5d7…`; only change: 630ebfd, additive). Applied
with `kubectl set image` on the Deployment only (generation 17 → 18, 1/1 Ready);
the file's Service, NetworkPolicies and Ingress were not reapplied. Verified in
the pod: `ensure_service_groups` and `revoke_service_groups` present, `/readyz`
200; no token used, no group changed. Rollback: `kubectl -n sso set image
deploy/identity-provisioner provisioner=forgejo.coulomb.social/coulomb/identity-provisioner@sha256:ffacd5d790a5b7800e0535f3aa95b7de52c7a68e4f575b9192c3e1622fa672c2`.
The pin in `sso-mfa/k8s/identity-provisioner/deployment.yaml` follows.

2026-10-09: ops-warden closed WARDEN-WP-0044-T01 and shipped `warden case plan`
(T02, message `06303d1d`); the accepted Forgejo case is refused until ops-mason
materializes it, and a materialized copy plans cleanly once
`keycape-client-secret-generate` exists. Its observation that `bind-env` carries
a literal `KEYCAPE_<CLIENT_ID>_CLIENT_SECRET`: the validator now renders a
`{param:env}` format spec (upper case, `-` as `_`, as `keycape-client.py`
derives the name) and rejects unknown specs, and follow-through and rollback
text is now checked for unknown placeholders too (patch-level, canon section 6).
The type keeps the literal until ops-warden's planner (`text.format`) renders the
spec as well; asked of ops-warden. 53 tests pass.

2026-10-10: ops-warden rendered `{param:env}` in `warden case plan`
(23ba584, message `c1df14b7`; other specs and `!conversion` refused). The type
adopts it as a 1.0 patch (canon section 6): `bind-env` and its rollback now read
`KEYCAPE_{client_id:env}_CLIENT_SECRET`. ops-mason reports the Forgejo case
rendered and built (message `1ebddae9`; CCR drafts handed to railiance-platform),
so T06 now waits on the platform apply, the mint-secret lane and the attended
register-client session.

2026-10-10: key-cape adopted the type's custody path (key-cape ab5cccd, message
`88a4559c`): `docs/forgejo-oidc-client-request.yaml` and KEY-WP-0038-T02 name
`platform/workloads/keycape/forgejo`, field `CLIENT_SECRET`, delivered as
`sso/keycape-forgejo-client` through `external-secrets-keycape-forgejo`, and the
`forgejo/netkingdom-oidc-client` proposal is withdrawn. This matches the type and
the Forgejo example; no change needed here.

2026-10-10: key-cape's type-composition review (key-cape 6038483, message
`8e5ed5f6`) found the composition sound with two blocking findings, taken as a
1.0 patch (stricter validation and documentation, canon section 6):
R1, redirect URIs now refuse `?` and `#` in the type pattern and in
`keycape-client.py`'s `REDIRECT` (KeyCape refuses `?` at startup and the single
replica would stop); R2, `allowed_scopes` always includes `groups`, the renderer
refuses `access_groups` without it, and `access_groups` now says the consumer
enforces the gate because KeyCape has none; R3, `mfa_required` "false" means the
per-user privacyIDEA policy; R4, register-client's justification records the
restart for every client, the quiet window, and that the rendered registry must
pass KeyCape's client validation before it. The accepted Forgejo case and both
examples still validate; 116 tests pass.
Open, needs a decision: R2's verification step that the consumer enforces the
group gate is a new verification step, hence 1.1 (minor), and the validator pins
cases to the exact type version, so the accepted 1.0 Forgejo case would need
re-pinning. Also open with key-cape: a KeyCape entry point that validates a
rendered client registry without starting the server; until then
`keycape-client.py` mirrors the redirect rule.

2026-10-10: founder go for 1.1, so no 1.0 run overlaps. The type adds
verification `consumer-enforces-groups`, and hand-over now requires the consumer
to enforce `access_groups` (minor, canon section 6; risk class unchanged, no new
acknowledgment; history in the type README). The accepted Forgejo case and both
examples are re-pinned to `oidc-browser-relying-party@1.1` before any apply; the
acknowledgment intake keeps `@1.0` as the record of what was acknowledged.
ops-mason's build and ops-warden's case-ref digest predate this and need a
re-render.

2026-10-10: ops-mason re-rendered the Forgejo case from `@1.1` (message
`22c6bba6`; plan digest `af1f4f81…`, no collisions) and rebuilt the bundle. The
rendered OpenBao objects are identical to 1.0, so the already applied
CCR-2026-0041 stands; 1.1 changes only consumer-side steps. key-cape shipped
`keycape check-config` (c528daa, message `a7b0aeff`): KeyCape's startup checks
without secret values, including the new duplicate-clientId rule.
`keycape-client.py --check-config <binary>` now runs it on the live and the
proposed registry, fed on stdin, and apply refuses a proposed registry that
fails; register-client's entry point names the option (1.1 patch). The deployed
image predates c528daa, so the binary is built from key-cape source for now.
Before the next KeyCape rollout (Forgejo register-client or the CCR-2026-0018
window), the attended operator runs `inspect --check-config` so the live
registry is checked too.

```task
id: NK-WP-0048-T07
status: wait
depends_on: [NK-WP-0048-T06]
blocking_reason: "The next types follow once the first one is proven end to end."
priority: medium
state_hub_task_id: "c3c9c3ae-4f23-5ba7-9e6f-e63bcf29b9f7"
```

### T07 — Queue the next types

From the T02 ranking, open the next type drafts (likely the confidential service
client and the workload KV read lane), each to be acknowledged on its own.
Record which existing hand-built instances get adopted as cases, without
rewriting them in place.

```task
id: NK-WP-0048-T08
status: done
priority: high
state_hub_task_id: "8e519082-dc8b-5e46-b660-03d03e55dcf6"
```

### T08 — identity-provisioner: ensure service-access groups

The type's `ensure-groups` step calls `identity-provisioner groups ensure --groups
<list> --members <group=user+user;...>`. The provisioner already has idempotent
`_ensure_group` and `_add_group`, but exposes only tenant-scoped operations. Add
the general operation (create if absent, add the listed memberships, never
remove others), its rollback (remove only this case's memberships), tests, and
the ops-warden route for its service identity.

2026-10-08: done. `identity-provisioner/provisioner.py` gained
`ensure_service_groups` (`POST /v1/groups/ensure`) and `revoke_service_groups`
(`POST /v1/groups/revoke`), behind the existing service-token check. Names
matching `tenant*`, `netkingdom-*` and `lldap_*` are reserved; members must
already exist; everything is validated before any change; memberships are read
back before success; nothing else is ever removed; groups are deleted only when
asked and empty. 7 new tests (39 in the provisioner suite). The type's
`ensure-groups` step and its rollback now name these endpoints. The ops-warden
route for the provisioner's service token belongs to WARDEN-WP-0044-T01.

```task
id: NK-WP-0048-T09
status: done
priority: high
state_hub_task_id: "0ee12dfc-8260-5038-bd39-cc7fb4113be0"
```

### T09 — The `keycape-client` entry point and its declaration

Planned first as key-cape KEY-WP-0038-T04. Moved here because the machinery a
live registration needs already lives in net-kingdom, where KeyCape is deployed
(`sso-mfa/k8s/keycape/`): the contained kubectl, the cluster-identity pin and the
issuer pin. A copy in key-cape would drift.

2026-10-08: done. `sso-mfa/k8s/keycape/keycape-client.py` derives the KeyCape
client record from a case (`client-from-case`). It adds or removes exactly one
client with the existing safeguards (byte-preserving merge, exact match or
refuse, shared-reference refusal, compare-and-swap, read-back, contained errors),
requires the Deployment env binding first for confidential clients, and requires
the case acceptance record for `apply` instead of a per-client founder decision
id. 15 tests, including a regression showing that the vergabe case reproduces
its hand-built registration exactly; all 71 helper tests pass. Declaration
`capabilities/playbooks/net-kingdom.keycape-client-registration.yaml` (contract
PASS, `maturity: implemented`) replaces key-cape's draft. The type now pins it,
and for confidential clients adds `deliver-secret` (delivery from custody into
the sso namespace) and `bind-env` (Deployment env), attended in the same session
as `register-client`. The per-client helpers stay until their clients are
adopted as cases.

2026-10-10 — NetKingdom owner review of the execution seam, source only:
ops-warden/docs/security-case-execution-proposal.md SHA256 78ab1439e4c5beb205aeccf15b1b2e706511be55ab195c5a3aa682f326c7c77c
was reviewed against INTENT, SecurityCase canon, acknowledged type1.1, current
case and owner source. docs/security-case-execution-owner-review.md supports
the closed adapter/receipt direction with explicit versioned schema, structural
materialization provenance, unchanged caller/session boundaries, safe CAS
reconciliation and exact-created-object rollback requirements. This is owner
review, not approval. Supporting file digests are in the adjacent sources JSON.

Current source exposes a circular wait: ops-mason T03 waits on mint while Warden
and the broker refuse a non-materialized case. Structural acceptance therefore
needs its own native pinned receipt before mint, not a successful mint before
materialized. The platform broker additionally pins type1.0 while Forgejo is
accepted at1.1; platform-owner alignment and tests are required, not an unchecked
version bypass. T06 remains wait. No case/type lifecycle status, grant, authority
or live runtime changed by this review. A proposed informed-decision would
adopt the source contract only; native materialization and execution acceptance
stay distinct.

Final proposal re-review: staged Warden proposal SHA256 78ab1439e4c5beb205aeccf15b1b2e706511be55ab195c5a3aa682f326c7c77c
adds only the explicitly local reducer appendix, not an accepted wire protocol
or executor. The exact-final review and sources JSON supersede the earlier
1035f049... proposal review. Platform staged the exact type1.1 broker correction
with ten fake-backend subprocess checks (script SHA256
9576cd211cfc06526ec4cc9bd0a5b06f0f59287a68931bb2d6e3592a951ddc7d).
The next source work is the versioned closed adapter/receipt/materialization
contract, not another relaxation of the broker's status gate. Native work still
requires installed broker provenance, structural reader receipts, exact new
materialized-case bytes, caller grant/session revocation and verified execution.
No installed/native acceptance is inferred from that staged correction.

2026-10-11 — new source-only type1.2 architecture acknowledgment:
Confirmed native founder ACCEPT of infd-20261011-forgejo-type12-source-r1@1
at06:47:39Z, approval8e9332c8-754a-420c-b8b7-445cb4fe6765. Exact frozen packet
0b377d647ca748cd03c92f9792396412048055d4bde9f7bb06ff6d413f7acd68 and submission
585cee731de30d2b37883a57abfd9eb32f023ee694e8d911989109d5caf80569; native receipt
retained in docs/evidence/2026-10-11-forgejo-type12-source-acknowledgment/.
The complete acknowledged1.2 source package is isolated under
security-case-type-revisions/1.2/oidc-browser-relying-party; default loader and
current accepted1.1 case remain untouched because load_type is single-version.
The new proposed1.2 case is NOT accepted/materialized/live and has no acceptance
fields. Actual reference-migration design approval is a separate Apps owner
receipt; no1.1 approval is inherited. Source-recording review/test proof and exact
next gates are in docs/forgejo-type12-source-acknowledgment-followthrough.md.
T06 remains wait and T07 remains its existing dependent task; type is NOT
implemented and no first-case/native grant or broader workplan finish is asserted.

2026-10-11 — bounded explicit version selection source preparation for T06:
The validator now offers optional exact version_roots/--version-root VERSION=PATH
selection with duplicate/mismatch/traversal/symlink refusal and no fallback. The
default1.1 source and historicalacceptedcase hashes/ordinaryrefusal/preparation
render stayunchanged; actual1.1 is acknowledged, not implemented. Acknowledged1.2
ordinaryadmission also refuses; optionalpreparation remains nonexecuting.43local
validatorchecks pass and original1.1 CLIrenderbytes match exactly. Makerproof:
docs/evidence/2026-10-11-explicit-type-version-selection/source-integration-validation.json.
FormalNetownerreview/reference-source integration and all existing native adapter,
provenance/broker/materialization/attendedproof gates remain; T06 stayswait, no
new residual or firstcase/typeimplementation/native grant is inferred.


2026-10-16 — isolated source preparation for T06 (canonical integration pending):
Added `check-evidence` to the existing validator CLI. It reuses exact version
selection and ordinary type/case/render admission, checks actual Mason structural
request source/inventory/projection pins and reads actual Warden journal rows
read-only. Successful rows require exact type-declared descriptors and attempt/
revocation fingerprints. Existing acknowledged types without machine descriptors
remain plan-only. The compatibility result never claims authenticated native
completion: Request and journal fingerprints are not portable native evidence.
Owner authentication/status binding and actual native gates remain unresolved;
T06/T07 statuses and both type/case lifecycles remain unchanged. No dispatch,
credential access or authority expansion is introduced. See tool README for the
existing owner protocol boundaries and source integration tests.

Concrete next Warden owner integration: current `Executor.admit` refuses every
case parameter whose sensitivity is not public/operational, so browser
`redirect_uris`, `allowed_scopes`, `client_type`, `mfa_required` (and 1.2
`consumer_namespace`) fail after otherwise prepared admission. The sensitive
classification is a security-change review level, not secret content. The
Warden owner must validate through NetKingdom's existing exact-version case/type
validator, then project only descriptor-declared safe target metadata; do not
remove the credential/value refusal or copy complete parameters into journals.
Actual owner descriptor acknowledgment and independent evidence admission remain
prerequisites, with normal admission still refusing unimplemented types.


## T10 — Source compatibility checker for Mason and Warden evidence

```task
id: NK-WP-0048-T10
status: done
priority: high
state_hub_task_id: "4b06ca97-c1e4-5680-bf22-ab75ab1646a4"
```

Completed 2026-10-16 in published source `c31d21ba`. The read-only
`check-evidence` command checks the actual Mason structural request and Warden
journal source protocols against an exact version/render. Independent source
review passed 103 tests at that checkpoint. It always reports native completion
false and execution prohibited. Fixtures and local journal fingerprints do not
provide authenticated owner/native evidence. This records completed source scope
previously described under T06; it does not duplicate native acceptance credit.

## T11 — Validate the optional machine-descriptor pin contract

```task
id: NK-WP-0048-T11
status: done
priority: high
state_hub_task_id: "ef2a0119-f59e-5e45-9519-115da9df7b85"
```

Completed 2026-10-16. Reproduced the JSON schema rejecting a valid
`execution_descriptors` map while the Python validator accepted malformed maps.
The schema now supports the existing Warden optional exact-byte pin map; semantic
validation binds keys to declared steps, bounds the map and checks digest shape.
The direct evidence API refuses malformed maps before accessing them. Twenty-one
new checks and all 127 affected source checks passed. No current type/case bytes,
status, owner descriptors or native acceptance changed. Source evidence and
publication receipt: Custodian
`docs/changes/netkingdom-target-2026-10-16/browser-acceptance-next/`.

## T12 — Prepare NetKingdom-owned attended step adapters

```task
id: NK-WP-0048-T12
status: todo
priority: high
state_hub_task_id: "7ce4eee3-cd47-55a9-bfa5-4f1835d4d6fd"
```

Owner: NetKingdom. Retained scope: prepare bounded attended Deployment env,
config verification and restart/readiness steps around the existing helpers.
Next action: identify exact current entry points and their revision/readback
contracts, then implement fixed source-pinned callbacks or receipt-only attended
steps. Completion requires positive/refusal checks, exact target/source pins and
uncertainty behavior without acquiring user-engine runtime identity or exposing
registry/client secrets. Existing caller/session, cluster/issuer, CAS/readback
and status-only output boundaries remain mandatory.

2026-10-16 task refinement: the original broad adapter scope is split into T13
(completed registration source candidate), T14 (group creation provenance), T15
(durable registration recovery) and T16 (descriptor/artifact selection and owner
review). This preserves T12's task ID and Hub UUID while giving each remaining
outcome an owner and completion gate. T06 now names every source dependency;
its native/reference lifecycle acceptance scope remains unchanged. Completion
of any source task supplies no Warden, Platform or Apps assent and no native
execution authority.

## T13 — Prepare the fixed registration adapter source candidate

```task
id: NK-WP-0048-T13
status: done
priority: high
state_hub_task_id: "b264c10b-565c-5d29-b788-03cf5fbd2883"
```

Completed 2026-10-16, source implementation and synthetic fixture proof only.
Added an injection-only adapter around actual `keycape-client.run`: fixed
case/principal/human/lane/source/acceptance binding, existing CAS/readback,
metadata-only owner evidence and per-attempt parent-session revocation boundary.
The descriptor source composition pins the exact checker executable digest
alongside adapter/helper source. Creation-only rollback authenticates its
creation receipt and refuses changed postcreation UID/resourceVersion;
pre-existing clients confer no rollback authority. Missing readback, evidence,
revocation or process-local creation provenance cannot become retry-safe success.

Evidence: Custodian `docs/changes/netkingdom-target-2026-10-16/browser-adapter-next/`
R3 candidate, explicit patch and postimages, retained R1 rejected custody finding
and R2 revision, and 36 passing source checks (16 new, 20 existing). Three checks
use actual Warden kernel admission/dispatch/rollback with published Warden
`d97652a3bb5fa733b063d71a1b74068efcccf572` producer pins unchanged. Owner trust,
descriptor/type/case and Kubernetes objects remain synthetic fixtures. This
records a completed source deliverable, not independent review, publication,
type acknowledgment, exercised procedure, identity admission or native browser
acceptance. Independent root source review: Custodian browser-adapter-next/coordinator-review-r3.json. Source integration and normal publication/synchronization receipts: Custodian browser-adapter-next/canonical-integration.json. These receipts supply no other-owner assent or native authority.

Source publication: `0775c2bfa41af236301155bb5a69bbf088bdf823`, independently
reviewed by the Custodian coordinator before integration. The required initial
projection and ordinary UUID writeback both completed with zero failures and
zero automation errors. Generated-only brief commits were refused by the normal
hook; exact pre-existing staged/working brief bytes were restored. The separate
canonical integration receipt retains those refusals and projection evidence.
Native admission and all other-owner acceptance remain open.

## T14 — Admit governed group rollback against native owner boundaries

```task
id: NK-WP-0048-T14
status: progress
depends_on: [NK-WP-0048-T20, NK-WP-0048-T21]
priority: high
state_hub_task_id: "fce95440-a25b-5f27-8ef3-a3d182099113"
```

Owner: NetKingdom. Next action: bind the completed T21 PostgreSQL journal to
an admitted dedicated CNPG consumer through the existing `rapp-postgres`
database-per-consumer contract. Identify the actual deployment/package and
custody owner, admitted placement/capacity and role/renewable lease/TLS target;
implement the independently retained exact content checkpoint and authoritative
tenant/case/source/principal callbacks before enabling governed group operations.
The injection-only backend and fixture callbacks do not supply production scope
or authorization. Existing service tokens, caller-supplied JSON and provenance
receipts confer no new grant. The default production endpoint remains disabled
without the trusted integration. SQLite remains local/disposable only; any
production exception requires explicit scoped governance under ADR-0015.

2026-10-17 refinement: T20 retains the historical durable SQLite source and
actual disposable LLDAP proof. T21 extracts the completed PostgreSQL source and
bounded real disposable SQL compatibility proof. T14 remains `progress` for
native CNPG admission/recovery, protected owner integration and exclusion of all
external LDAP writers. Separate source publication/projection and independent
cleanup receipts belong to coordinator follow-through; no missing native assent
is represented as received.

Completion still requires independently reviewed/admitted ADR-0015-conforming
CNPG storage (or an explicit governed exception), backup/lost-store recovery and
migration of legacy requests; a native
writer-isolation/fencing or provider CAS/version contract that closes external
writer races and remove/re-add ABA; exact case/source/principal and current
native object bindings for protected callbacks; and actual per-attempt evidence,
revocation and readback/refusal verification through the existing owner paths.
Only independently retained confirmed creations may be undone. Every
pre-existing/shared group and membership must remain preserved; unresolved
intents never expire into retry authority. The disposable fixture supplies none
of these native custody, authorization, owner-acceptance or production claims.

## T15 — Define authenticated recovery for lost registration process state

```task
id: NK-WP-0048-T15
status: todo
priority: high
state_hub_task_id: "1b825705-4848-5edc-832f-b1622cd9bdb7"
```

Owner: NetKingdom for registration state, with Warden owning kernel evidence
admission. Next action: select the exact existing owner evidence holder/readback
contract and define authenticated lookup for the original creation attempt.
Completion requires restart/lost-receipt fixtures that bind case/source/principal,
original created-only provenance and current native revision before any undo;
uncertain outcomes remain non-replayable. The T13 adapter deliberately refuses
post-restart rollback and installs no reconciliation callback. A new broker or
unverified JSON receipt is not a recovery contract. Actual evidence-holder
acceptance and attended recovery remain separately authorized native work.

## T16 — Select and review the actual machine descriptor composition

```task
id: NK-WP-0048-T16
status: todo
priority: high
state_hub_task_id: "200bf500-5084-5ae7-be26-3033fe1f296e"
```

Owner: NetKingdom for descriptor/type source and exact checker artifact; Warden
independently owns global provenance/admission. Next action: obtain review of the
exact checker source/artifact proposed by T17 and identify native owner evidence/
session callback contracts, then assemble the complete step descriptor set and
a proposed explicit type version selection. T13's historical descriptor retains
its synthetic checker; T17 supplies an artifact-bound proposal, not admission.
Completion requires separate owner source review of exact machine descriptors,
artifact/source/config pins, normal type/version admission and independent
evidence-verification boundaries. Warden d97652a refreshes producer validator
pins for NetKingdom 015694ee; it does not attest T13's new adapter. Pinning any
machine descriptor into an acknowledged type is a separately reviewed source
change. Platform owns structural/mint/broker/delivery evidence; Apps owns actual
consumer login/group enforcement. Neither their assent nor native acceptance is
supplied by this task.

2026-10-16 source checkpoint: T19 below extracts the completed per-attempt
executable snapshot implementation and reviewed fixture proof. T16 retains
native executable storage/custody, exact owner evidence/session callbacks,
complete descriptor composition and ordinary type/version admission. The
successor descriptor/source pins are proposed only; the earlier descriptor
history is preserved. T14 group provenance and T15 durable restart/recovery
remain independent open requirements. No native acceptance is counted here.

## T17 — Build and exercise the exact offline config-checker artifact

```task
id: NK-WP-0048-T17
status: done
priority: high
state_hub_task_id: "8c3e4baa-a218-5996-af9e-fe1153a46dac"
```

Completed 2026-10-16, isolated artifact/source fixture proof only. Built the
existing `key-cape/src/cmd/keycape` command from published clean source revision
`3c71226a1a7b8a2bdb710e9c497e1a1f7f8a3987`, source tree
`f0010b9cab17dbc3a787628a29657d98b900320f`, with cached Go 1.23.0,
CGO disabled, Linux/amd64 v1, trimmed paths, disabled VCS stamping and offline
read-only module selection. The full retained closure pins 108 source files,
9876 toolchain files and ten used module source directories. Three isolated
build locations produced identical 13633409-byte artifacts:
`sha256:be37738fe1c08259fb4f9ca0daf0f9301bfffc237e5bb38435525dfbf11ccff5`.

Eight disposable checks use the actual executable and maintained helper/kernel,
with complete synthetic configs and a minimal environment: add/created-only
removal, normal pinned producer admission, rollback, lost-readback uncertainty,
stale revisions, substituted checker refusal, invalid startup config and missing
secret-env refusal. No checker stub or producer-pin override remains. The
artifact-bound descriptor proposal is
`sha256:76b1153d188062ece53fcd5e8d35290fdbfd02a69dd79d4b001fc42781afcdc3`.
Evidence: Custodian
`docs/changes/netkingdom-target-2026-10-16/browser-checker-artifact-next/`.
The first two build attempts succeeded but their closure collection failed;
both failures are retained, as is refusal of the older incomplete synthetic
fixture. No production config, credentials or cluster mutation was used.

This extracts completed build/fixture scope from T16 without claiming KeyCape
owner endorsement, native checker selection/storage custody, acknowledged type
integration, native evidence/session callbacks or attended browser acceptance.
T16 retains exact owner review and composition/admission scope. The attended
registration procedure remains unexercised.

## T18 — Make checker input and executable selection deterministic

```task
id: NK-WP-0048-T18
status: done
priority: high
state_hub_task_id: "9c45da71-fb4c-5f20-9065-90f448f37c64"
```

Completed 2026-10-16, isolated source correction and real-artifact fixture proof
only; independent integration/publication remain separately recorded. KeyCape
source review identified non-secret `KEYCAPE_ACCOUNT_PORTAL_URL` and
`KEYCAPE_BROWSER_LOGOUT_URL` overrides in config.Load. The original helper
inherited ambient environment, allowing valid supplied configuration to be
rejected or invalid supplied URL configuration to be admitted. A bare relative
checker filename could also be replaced by PATH lookup after the adapter had
hashed the selected local file. All three counterexamples are retained against
the unchanged predecessor.

The existing helper now resolves the selected checker file to an absolute path
and passes an explicit empty subprocess environment. Three opt-in real-checker
regressions verify both URL-verdict directions and relative-file selection under
a non-checker PATH decoy. All 39 affected checks pass with the exact supplied
artifact, as do 11 real-artifact helper/kernel scenarios. Source evidence:
Custodian `docs/changes/netkingdom-target-2026-10-16/browser-checker-environment-r2/`.
Historical descriptors remain retained; the helper correction changes successor
source-composition pins and requires their separate review/admission. An absolute
path is not immutable custody and does not eliminate file replacement races;
T16 retains native executable storage/custody and callback/type admission.
No production configuration, credentials or cluster mutation was used.

Source integration and publication: `916ba4b33e03179274d03b78bd898a4843e8653f`, following independent
Custodian source review in browser-checker-environment-r2/coordinator-review.json.
The ordinary protected projection pass assigned T17/T18 UUIDs with zero failures
and zero automation errors; generated-only brief commit refusal is retained.
The exact prior staged/working brief bytes were restored. Final source/publication/
projection evidence is retained in Custodian browser-checker-environment-r2/
canonical-integration.json. These source receipts do not admit native executable
custody, owner callbacks, acknowledged type changes or browser identities.

## T19 — Bind checker execution to verified immutable attempt bytes

```task
id: NK-WP-0048-T19
status: done
priority: high
state_hub_task_id: "f50fdcd1-00be-5c3c-be0c-345aa2126032"
```

Completed 2026-10-16, independently reviewed source implementation and synthetic
fixture proof only. The registration adapter now opens a bounded regular checker
artifact without following its terminal symlink, verifies the reviewed digest,
and snapshots the exact bytes into a write/grow/shrink/seal-removal-sealed Linux
memfd before helper access. Live and proposed config checks execute that same fd
through procfs with the existing empty environment; the fd closes on completion
or failure. Unsupported snapshot/seal/exec policy refuses through the existing
parent-session revocation and uncertainty boundaries. The earlier unbounded
pre-snapshot path read was removed during independent review; actual adapter
symlink/FIFO/oversize refusal tests verify helper access never starts.

Evidence: Custodian `docs/changes/netkingdom-target-2026-10-16/browser-resume-r1/`,
including exact base/preimages/postimages/patch, proposed descriptor, coordinator
source review and canonical integration receipt. Independent Custodian checks
passed 49 affected source tests and 12 actual Go checker/helper/kernel scenarios;
the sets overlap and are not separate acceptance counts. Existing published
Warden d97652a producer pins remain unchanged. The exact KeyCape artifact remains
`sha256:be37738fe1c08259fb4f9ca0daf0f9301bfffc237e5bb38435525dfbf11ccff5`.
The predecessor path-replacement refusal and the first expanded reopened-fd test
failure are retained; the latter fixture was corrected to restore write mode
before independently proving seals block reopened-fd writes and shared mmap.

This source task adds no native owner custody, callback, evidence trust, type
acknowledgment, credential access, consumer sign-in or rollback acceptance.
Executable memfds, Linux seals and usable procfs are explicit runtime dependencies
and still require native admission. This is an extracted source outcome from
T16, not completion of T16; T12/T14/T15 and all external-owner gates remain open.


Publication verified by the coordinator after the source worker usage limit:
`8790dd18ae83c4651ef3d1d923312a118a5b436b` is the actual remote main revision.
All five canonical source hashes match the reviewed packet. Normal synchronization
completed with zero failures/automation errors; all 19 task statuses match the
projection. T19's assigned UUID is retained. The original staged and working brief
bytes were restored and independently compared. The exact source, projection and
publication checks are in Custodian browser-resume-r1/canonical-integration.json.
Native acceptance remains open under T16 and the other named owner tasks.


## T20 — Prepare durable group provenance and prove disposable LLDAP compatibility

```task
id: NK-WP-0048-T20
status: done
priority: high
state_hub_task_id: "863fcbdb-e246-51ba-87ca-a2f44d01d6f8"
```

Completed 2026-10-17, source candidate preparation and bounded synthetic
compatibility proof only. This extracts evidenced implementation/fixture scope
from T14; T14 retains native storage, writer isolation, protected callback and
owner-acceptance gates. Source integration/publication and State Hub projection
are pending the coordinator's separate receipts; no revision or task UUID is
invented here.

The existing identity-provisioner group ensure/revoke interface now has a
reviewed candidate for a bounded owner-private SQLite journal, proven only in
local/disposable use. The initial storage instructions wrongly suggested a native
persistent-volume path; source review against accepted ADR-0015 corrected them
before publication. Production SQLite-on-a-PVC is not admitted: T14 retains
CNPG/rapp-postgres storage admission or a separately governed exception.
The local journal saves
complete prestate and intent before each attempted directory mutation, inventories
only confirmed new groups/memberships, binds original ensure idempotency and
correlation to the exact request digest, and refuses unsigned caller inventories
or legacy broad revoke requests. Replays never recreate removed state. Missing,
replaced or untrusted storage and missing intent rows stop execution; definite
before-mutation refusals retain evidence without stranding unrelated valid work.
Interrupted writes/readback remain unresolved across reopening. Revoke validates
all targeted native ids and complete member postimages before destructive work,
removes only confirmed creations and never deletes a preexisting empty group.
Server/configuration wiring, image-copy closure and owner storage/request
migration instructions are included. The existing endpoint's caller
authentication remains the existing boundary, not a new grant.

The candidate was prepared against NetKingdom
`878abc93588e12b6da1bac055cfbb3abb8d7adf1`; exact source patch digest
`sha256:1f3f3ca94c46dadac8f24896053d4d23179de795e50a24298a0cfbcc46cca30e`.
The independent Custodian review passed all 65 provisioner checks. The earlier
62-check candidate, two independent-review findings and subsequent fixes remain
retained rather than overwritten. Seven additional fixture-plan checks exercise
the preparation boundary; these counts overlap the same candidate and do not
measure task-completion percentage or independent acceptance.

A separately reviewed bounded fixture then exercised the exact candidate against
cached pinned disposable LLDAP and Python images, through actual GraphQL. All 14
closed compatibility/containment oracles passed: exact creation inventory,
separate-process SQLite reopening/replay/revoke, complete preexisting directory
preservation including empty/shared groups and memberships, refusal of shared
membership drift, retained uncertainty after a lost response following a real
addition, and refusal of unknown/cross-case receipt bindings. The fixture used
only newly generated synthetic users/passwords/key material in private tmpfs,
with a network-none anchor and shared loopback helper, no host ports/mounts,
no image pulls and no production or actual credential contact. The coordinator's
safe receipt reports both exact owned containers absent after cleanup; final
independent cleanup verification remains the coordinator's integration duty.

Evidence: Custodian
`docs/changes/netkingdom-target-2026-10-16/group-provenance-r1/`, including exact
source base/pre/postimages, patch/manifest, independent review and retained
refinements. Native fixture manifest
`sha256:8a0514f5f14137355972400cd4852f48c892316d28e3abd3c86635d7cb81eef2`;
retained safe result `native-fixture/execution-result.json` (original
`/tmp/nk48-groups-native-review-5pywb19z/result.json`). The source task supplies
no native owner assent, production rollout, protected session/evidence callbacks,
ordinary type admission or consumer login proof. LLDAP has no conditional
membership mutation or generation token here: exact readback cannot exclude
ABA or a concurrent external writer, so T14 remains open.

2026-10-17 verified source follow-through: source commit `4f65e41` passed normal
local scanning and all 65 canonical provisioner tests. The actual disposable
LLDAP receipt passed all 14 oracles; independent verification on the pinned
daemon found both exact names and the run label absent. Source integration
includes the ADR-0015 documentation correction; the journal remains local-only.

T14 source preparation is underway. Twenty-three pinned public sources identify
NetKingdom implementation/reconciliation, Platform custody/database governance
and rapp-postgres consumer provisioning; the inspected runtime binding does not
assign identity-provisioner a package. The existing User Engine runtime token
and pod route are not a protected-case caller lane. The native proposal preserves
the same journal seam on a dedicated CNPG consumer and requires authoritative
scope binding, admitted role/lease/placement/recovery and all LDAP writers fenced.
The declared exclusive maintenance window is not mechanical exclusion of WebUI,
helper or lifecycle writers. Capacity placement and EUR cost/risk remain owner
decisions. No package, database, credential, rollout or owner assent was created.
Custodian evidence: group-provenance-r1/native-storage-proposal-r1; final source,
projection and cleanup review: resume-integration-r1/publication-verification.json.


## T21 — Implement the PostgreSQL journal and prove disposable SQL compatibility

```task
id: NK-WP-0048-T21
status: done
depends_on: [NK-WP-0048-T20]
priority: high
```

Completed 2026-10-17: bounded source preparation and actual disposable PostgreSQL
compatibility only. T14 retains native storage, custody, callbacks, authorization,
all-writer isolation and production admission. Coordinator source publication and
projection remain separate follow-through; no source revision or task UUID is
invented here. T20 remains the historical local SQLite/actual LLDAP proof.

The eight-path source candidate adds an injection-only PostgreSQL implementation
behind the existing owner journal seam, with explicit trusted tenant/directory/
case/source/principal/epoch binding and independent continuity callbacks. It
persists bounded intent/created-only receipts, exact-one version updates, durable
active fences, and a content-bound journal chain. SQL commits precede independent
exact-marker acknowledgment; lost replies/checkpoints poison continuation and
retain uncertainty. A counter-only checkpoint is explicitly insufficient for
divergent clones. Migration does not create database/schema/credential, admit an
epoch or clear quarantine; production HTTP activation is absent pending trusted
callbacks. Pinned psycopg package/image-copy closure is included, without claiming
an image build or native dependency admission.

Candidate base `6cb43c0cc55212363a5406b02db046e2384a7c40`; final eight-path patch
`sha256:5549d2982b61ff70aba4b3928fecd5639db0f1a732ab2f8630f678e7acf4d34d`.
All 82 source tests pass, alongside 13 fixture plan/diagnostic checks and two
source-backed snapshot regressions. Counts overlap the same source and fixture;
they are not additive native acceptance or completion percentages.

The independently reviewed R5 fixture ran the actual SQL migration/backend and
existing provisioner path against cached pinned disposable PostgreSQL. Directory
operations deliberately used the existing synthetic LLDAP test double. All
manifest-declared phase oracles passed: creation-only inventory and DDL refusal;
separate-process replay/revoke with complete preexisting/shared visible-state
preservation; foreign-context refusal; concurrent exactly-once reservation and
stale version refusal; actual database fork with divergent bodies at the same
counter and fresh-object rejection of the losing exact checkpoint; retained
uncertainty after checkpoint/directory/SQL response loss; and actual pg_dump/
pg_restore rollback refusal against the independently retained fixture checkpoint.
The checkpoint is fixture-only, not an admitted native owner evidence holder.

Only generated credentials/state were used, with two nonroot/read-only containers
on private loopback, bounded tmpfs/CPU/memory/PIDs/lifetimes, no host ports/mounts,
no pulls and no production contact. The safe receipt reports both exact owned
containers absent, with no daemon uncertainty. Independent final pinned-daemon
cleanup verification remains coordinator follow-through.

Failed R2/R3/R4 receipts and source revisions remain retained. R2's helper failure
subphase is unknown; R3 exposed a real backend bug where replacing nested receipt
objects lost the created ID held by ensure, fixed by acknowledging only the backend
version after commit/checkpoint. R4 verified that fix through actual SQL bootstrap,
then exposed fixture snapshot aliasing and empty deleted-group backing entries;
source-backed regressions corrected the snapshot without weakening complete
visible-prestate equality, and reject nonempty orphan membership. These refinements
belong to one trial, not extra native acceptance successes.

Custodian evidence: `docs/changes/netkingdom-target-2026-10-16/group-postgres-r1/`;
exact source/pre/postimages and retained predecessors; native R5 manifest
`sha256:d51f90816616848b93bcdc482e10a98d07061deb01eeed9e50700fc6e68b6051`;
`final-integration-r1/typed-execution-evidence.json` and byte-exact safe receipt
(original `/tmp/nk48-postgres-native-review-sguj7t8p/result.json`). No CNPG native
storage/HA/backup/lease/TLS/capacity/restore admission, authoritative protected
callbacks, real LDAP writer fencing/CAS/ABA exclusion, owner assent or production
endpoint activation is claimed. Those remain live T14 work.