access-engine/workplans/FLEX-WP-0024-decision-envelope-authenticity.md

---
id: FLEX-WP-0024
type: workplan
title: "Sign the decision envelope: the response channel is unauthenticated"
domain: infotech
repo: flex-auth
status: active
owner: claude
topic_slug: netkingdom
planning_priority: P1
planning_order: 240
depends_on_workplans:
  - FLEX-WP-0021
related_workplans:
  - FLEX-WP-0019
  - FLEX-WP-0023
created: "2026-09-06"
updated: "2026-10-16"
state_hub_workstream_id: "90577acd-6910-548d-a13e-1dbfdfb8ed27"
---

# FLEX-WP-0024 — Sign the decision envelope: the response channel is unauthenticated

Opened by `FLEX-DEC-2026-010`, from a question `secrets-engine` asked rather
than answering locally: **how is the response channel authenticated?** They
declined to choose a transport control for flex-auth's service on the grounds
that it is not a consumer's call. Correct, and it is why this is a flex-auth
workplan and not a mitigation in their repo.

The stance today: pins serve plain HTTP, `flex-auth.decision-record.v1` carries
no signature, and a responder that knows the package id and version — both
published — can return a well-formed `effect: allow` that passes every check a
consumer performs.

## The two things not to get backwards

**The digests do not authenticate the responder and cannot.** Every input to
them is either sent by the caller or published, so a forger reproduces
`request_digest`, `policy_package_digest`, and `registry_snapshot_digest`
exactly. Publishing more digests makes a forged envelope look *more*
authenticated, not less.

**Fail-closed does not cover this.** It protects against a PDP that is absent,
not against one that lies. An unreachable PDP denies; a lying PDP allows.

## 1. Publish the stance and stop publishing bare Service names

```task
id: FLEX-WP-0024-T01
status: done
priority: high
state_hub_task_id: "b8fad80d-5c44-5e26-a632-5096e0c0f3c5"
```

Owner: `flex-auth`.

**Done 2026-09-06.** `docs/decision-record-contract.md` states that digest
recomputation is not responder verification, because the contract is the only
document a consumer is guaranteed to read.
`docs/operator-caller-access-path.md` carries the reproduced DNS probe and the
channel analysis. The trailing-dot FQDN replaced the bare Service name in
`examples/secrets-engine/README.md`, `SCOPE.md`, and the `FLEX-WP-0021-T05`
handover note — a bare `*.svc.cluster.local` name resolves through
`search ad.binect.de` to one unrelated public host, and a name for a service
that does not exist resolves to the same address.

## 2. Choose the signature shape and key custody

```task
id: FLEX-WP-0024-T02
status: done
priority: high
state_hub_task_id: "5482f3cd-36cb-55e0-8c52-0ca2340ed4d7"
```

Owner: `flex-auth`; key custody is **not** ours to invent.

- Detached signature over the canonical envelope, preferred over transport-only
  mTLS for two reasons: it survives the transport, so a decision audited later
  is still attributable, where mTLS authenticates a connection that no longer
  exists; and it composes with `FLEX-DEC-2026-009`, since a record that names
  its authenticated caller *and* carries flex-auth's signature is an audit
  artifact where either alone is a claim.
- Decide the canonical form to sign. The existing canonical-JSON rules from
  `docs/canonical-request-digest.md` are the obvious basis, but the signed
  material is the **envelope**, not the request, and it must exclude the
  signature field itself. Write it down before implementing: an implicit
  canonicalisation rule is what `FLEX-DEC-2026-007` had to correct.
- **Do not mint a key in this repo.** Route custody via
  `warden route find "decision envelope signing key"` and OpenBao, per the
  estate credential-routing rule. flex-auth decides the shape; it does not own
  key material.
- State key rotation and how a consumer discovers the current public key, since
  a consumer that cannot rotate trust has a pinned key that will one day be
  wrong.

Gate: the canonical form, the algorithm, and the custody path are written down
and reviewed before any signing code exists.

**Done 2026-09-14.** Written in `docs/decision-envelope-signature.md`: detached
Ed25519 over `json.Marshal` of the envelope with `signature` omitted; unsigned
is `{mode: unsigned}`; production key lives in OpenBao, not this repo
(proposed catalog id `flex-auth-decision-envelope-signing-key`). Test-only seed
is documented and is not the custody path.

## 3. Implement signing and verification

```task
id: FLEX-WP-0024-T03
status: done
priority: high
state_hub_task_id: "041612ea-1be4-5997-8c32-49f8bcc50855"
```

Owner: `flex-auth`.

- Add the signature to `schemas/decision_envelope.schema.json` as additive, and
  state in the contract that `request_digest` is unaffected so no consumer
  re-pins — the same note that was needed for v2 and for
  `approval_binding_digest`.
- Emit unsigned when no key is configured, and say so in the envelope rather
  than omitting the field silently. Absence stated, not implied — the
  `pdp_digest` precedent.
- Ship a verification example and a fixture pair in
  `examples/secrets-engine/replay/`: a genuine signed envelope and one whose
  payload was altered after signing, so a consumer's verifier is tested against
  a real failure rather than only a real success. **A verifier that has never
  seen an invalid signature is untested**, which is the lesson from every
  fixture suite in `FLEX-DEC-2026-008`.

Gate: an altered envelope fails verification, an unaltered one passes, and a
consumer can perform both from the fixtures alone.

**Done 2026-09-14.** Additive `signature` on the schema and envelope. Engine
emits `{mode: unsigned}` when no key is configured. Fixture pair:
`examples/secrets-engine/replay/decision_rotate_signed.json` verifies,
`decision_rotate_signed_tampered.json` does not, same `request_digest`. Public
key in `keys.json`. Tests in `internal/sign`.

## 4. Report the gap to gate-house

```task
id: FLEX-WP-0024-T04
status: done
priority: medium
state_hub_task_id: "82d6b75e-e2c6-5e3d-a897-fbdfb7c9094e"
```

Owner: `flex-auth`.

§6.4 obligates the PEP to act on a decision and §9.3 splits degraded mode by
owner, but **nothing in the standard requires a PDP's response to be
authenticable**. So a fully conformant PEP can trust a forged allow while
meeting every published obligation. That is a gap in the standard, not only in
this implementation, and it belongs in the outstanding v0.8 assent round
alongside the `FLEX-DEC-2026-009` finding.

## T04 closed — 2026-09-07

**Done.** Reported as `FLEX-DEC-2026-011` F1, ranked blocking for
calling the §6.4 PEP obligations complete and explicitly not blocking anyone's
adoption of v0.8 — the condition already exists under v0.7 and is not created by
this version. §16 already carries the same observation one layer up, about the
standard's own publication path having no digest or freeze discipline; F1 is
that observation applied to the artifact the standard regulates.

## 5. Wire owner signing into the decision pin and retained deployment charts

```task
id: FLEX-WP-0024-T05
status: done
priority: high
state_hub_task_id: "cfe4d275-9f02-5090-8391-06eea59d45c5"
```

2026-10-16 source follow-through: NetKingdom target Secrets Engine R2 owner review
found that the completed T03 signature package was used only by fixtures. Add
explicit mounted seed/kid configuration for `serve`, sign the final envelope before
recording, and expose inert existing-Secret mounts in both retained/new charts.
The API/sign verifier sources remain byte-identical. Local source validation and
review are in the Custodian `docs/changes/netkingdom-target-2026-10-16/access-verifier-review/`.
This is source preparation only, not signer admission or deployed acceptance.

## 6. Admit custody, consumer trust, exact packages and native signed decisions

```task
id: FLEX-WP-0024-T06
status: progress
priority: high
state_hub_task_id: "7f785b52-bf62-5884-8d34-670d70bc9023"
```

Wait on Platform/OpenBao key custody, independent public-key and issuer/policy-pin
admission, exact image and existing-Secret values rollout, owner-bound caller/tunnel,
and Secrets Engine verifier installation. Follow `docs/decision-envelope-signature.md`
for overlap/removal semantics. Retain signed producer record/consumer acceptance,
forged/altered/unsigned/wrong-pin/expired denial and rotation-removal receipts;
then update actual deployed status. No signer or runtime trust is admitted by the
source review. This task owns the producer-side native gate; SECRETS-WP-0014 owns
consumer admission/enforcement. The old fixture completion never established this
gate, so the workplan is reopened rather than claiming it already passed.


2026-10-16 owner native preparation: T06 is `progress` because Access still owns
signer workload declaration, artifact/provenance and source/provider-join preparation;
no unqualified cross-owner wait is claimed. Read-only actual release metadata and
bounded proposal are in the Custodian
`docs/changes/netkingdom-target-2026-10-16/access-native-preparation/`.
The real Platform custody route `flex-auth-decision-envelope-signing-key` exists but
is draft/nonresolvable; no production kid/public fingerprint, Secret projection or
custody receipt is admitted. Coordinate that named commitment through CUST-WP-0080
instead of inventing a Platform task/approval. SECRETS-WP-0014-T03 owns installed
consumer native verification and must not form a circular blocking dependency.

The first proposed native trial only signs the unchanged lifecycle package on the
existing retained release; `execute-ccr` remains denied. Current service-pilot
`runtime_admission := false` remains mandatory: the source ServicePilotInputs is
not yet wired into HTTP, and public service_actor_binding never authenticates the
caller. Later activation needs a complete reviewed provider join, exact policy and
operation admission, artifact digests, owner execution/rollback proposal and native
receipts. Preparation/local synthetic success grants none of those permissions.


2026-10-16 source-only native preparation adds the existing Secrets PDP ServiceAccount
to `tenancy.yaml` with evidence in `docs/native-signer-workload.json`, makes the
retained `flex-auth` audience explicit, pins CI archive fetch to the full source
revision, and adds offline native signer receipt verification using the existing
Secrets verifier. No key, trust, image publication or live activation is admitted.
The managed workload's missing rapp declaration remains explicit migration debt;
this source binding does not fabricate a package declaration or raise its grade.


2026-10-16 separate B1 source follow-through wires ServicePilotInputs through actual
HTTP/TokenReview into a signed-DENY-only engine entry. Generic `execute-ccr` calls
refuse across every policy, including the old separately named positive fixture;
that proof stays pinned to ec21518e. Original context/submitted digest stay unchanged.
Actual disposable serve/TLS TokenReview/HTTP signed-DENY and bypass/refusal tests are
source evidence only. B2 depends on independently reviewed Approval fact contract
and canonical vectors; B3 additionally needs exact source/normalization numeric
representation and full trusted provider joins. No current policy false gate,
production trust, native deployment or protected effect is enabled by B1.

2026-10-16 B2/B3 source successor preparation (separate from frozen B1): the
isolated v2 source path now joins the fixed authenticated Approval reader,
unchanged Approval-owner canonical/fact helper, original/fresh correspondence,
raw numeric request/registry source retention, same policy evaluator and signed
source provenance. Disposable actual HTTP and production serve subprocess prove
synthetic signed ALLOW and fail-closed negative joins with exact response/record
parity. See `docs/trusted-service-task-source.md`. T06 remains `progress`: this
source proposal requires independent cross-owner review and full consumer/Platform
successor integration, then separately admitted runtime custody, source trust,
image provenance and rollout. No source-only proof establishes those runtime gates.


2026-10-16 B2/B3 R2 temporal correction (R1 retained): whole-stack review exposed
that R1's 15s ALLOW and default Approval freshness30 cannot fit the declared
minimum backendTTL30 plus existing cleanupreserve5. The v2 successor now derives
exact ALLOW expiry from the immutable operationdeadline, requiring all original/
fresh approval/freshness/caller bounds, a conservative source maximum120 and
remaining declaredTTL+5 after source I/O/evaluation. No consumer/native deadline
or cleanup guard is relaxed. Max structural budgets that cannot fit refuse.
The fixture declares backendTTL30/deadline60 and explicit source freshness120;
owner runtime source freshness configuration and full native integration remain
separate reviewed gates. See docs/trusted-service-task-source.md.

2026-10-16 B2/B3 R3 follows actual retained-registry integration failure: subject
enrichment changed evaluated approval-free digest from the original proposal.
The existing signed claim-free DENY now explicitly carries v2 approvalfree digest
(equal evaluated requestdigest), enabling producer-owned preapproval facts under
the already published enrichment contract. The final original proposal adds only
the issued claim; registry drift/refreshed facts cannot silently rebind it.
Retained-registry HTTP and production serve proofs and negative joins pass.
R3 preserves precise notbefore pending separate Approval-owner timestamp fidelity
repair; no rounding waiver, copied registry algorithm or empty-registry workaround.

2026-10-16 native admission prerequisite: fresh metadata confirms no namespace
PVC exists, and current outbox had neither OS writer exclusion nor a byte/startup
bound. Prepared existing-outbox hardening on exact canonical R3 base2b064ee:
nonblocking Linux flock held through Close/process lifetime before recovery,
64MiB source journal ceiling,2MiB record/atomic-batch bounds, streaming startup,
4KiB cursor, all-effect commit refusal and fenced unconfirmed rollback. Actual
process contention/crash release, capacity/prefix/restart and ENOSPC/fsync failure
proofs are part of the owner package. No new service, live mount, deletion or
custody grant; exact published successor, new PVC/PV retention/managed-package
owner disposition and independent consumer admission remain prerequisites.

2026-10-16 owner package extraction record: canonical local package
`rapp-flex-auth-secrets-engine` now exists at commit
`d52f811ed5ec0be27c9c367e71d476125b3228be`, tree
`d4fbad71bdc8ef870603526ccd1f53195d0322af`; authoritative declaration
`declarations/rapp.yaml` SHA256 `0b722431bc5162dd86f72b15f12424e149aca8f2a8790adf2c57e8511ea73b4c`.
Repo Manager exact local resolution under `/home/worsch` returns
`rapp-flex-auth-secrets-engine` / workload `flex-auth-secrets-engine` /
deployable `flex-auth-secrets-engine`, ownership `access-engine`, readiness
`draft`. The retained lifecycle PDP is the only rollback boundary claimed;
tenant/user/canary and retired `rapp-secrets-engine` remain excluded. Normative
focused family validation passes; existing unrelated global rapp-forgejo error
and reef projection warning remain visible. Package extraction keeps source
`2b064eee510640224274ccdc7cefceb1ae13162e` and current image
`sha256:05a03a8790c2210c48ea92391441c77ddf640d0cd32f5ec09838f5393171fcbd`
unchanged. Exact local source resolution is no central/private read-placement,
reef binding, custody, native rollout or consumer trust admission. Future signer
image/config adoption remains separate. T06 remains `progress`; no second work
item duplicates this bounded dependency.