secrets-engine/tests/test_decision_replay.py

490 lines
19 KiB
Python
Raw Normal View History

fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
"""Digest join verified against real flex-auth DecisionEnvelopes (FLEX-WP-0021-T03).
These replace a hand-maintained pin that was computed *including* the request
`id`. docs/canonical-request-digest.md excludes `id`, `policy_version` and
`caring_context` from the hashed material, and both real envelopes confirm it:
a digest computed over the old material matches no issued decision, which would
have failed closed against every correct allow.
Both fixtures are required. `provenance.input_claim_digests.context` appears
only when the request carries a non-empty context, so a validator asserting it
is always present passes `destroy` and fails `rotate`.
"""
import json
from datetime import datetime, timedelta, timezone
from pathlib import Path
import pytest
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
from types import SimpleNamespace
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
from secrets_engine.authorization import (
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
REQUEST_TENANT,
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
approval_binding_digest,
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
build_action_request,
digest_material,
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
digest_material,
request_digest,
validate_decision_envelope,
)
feat: adopt the owner-documented PDP access path, and prove it live Adoption asked for by flex-auth (FLEX-WP-0021-T05) and glas-harness (GLAS-WP-0015), plus the first real decision this engine has obtained from the deployed pin -- which found a defect the fixtures could not. ACCESS PATH. require_supported_pdp_address refuses in-cluster Service names and any non-loopback host. This is no longer a unilateral call: the owner path is documented as loopback kubectl port-forward over the authenticated Kubernetes API, which is what authenticates the responder transitively (FLEX-DEC-2026-010). A Service name from a workstation does not fail, it resolves through the DNS search suffix to an unrelated public host, and since decision records carry no signature, a responder knowing the published package and version can return an allow that passes every check we make. Fail-closed protects against a PDP that is absent, not one that lies. The guard runs before the token is read, so a misdirected request cannot leak it; a test pins that ordering. LIVE PROOF. Minted a 10-minute TokenRequest token (audience flex-auth, SA secrets-engine/secrets-engine, mode 0600 outside the worktree, shredded after), forwarded to the named pod, and sent a real CheckRequest for glas-claude-agent-dev-anthropic. Result: allow, catalog_lane_policy_matched, served by v2 (sha256:bd11c5fe...) -- so the redeploy flex-auth flagged as outstanding has landed and the pin no longer serves the tenant-blind v1. Our tenant fix is confirmed against the real service: binding.tenant is tenant:platform. THE DEFECT IT FOUND. The evaluator enriches from its registry before hashing -- subject gains attributes and tenant, resource gains tenant -- so binding.request_digest is over material we never sent and cannot reproduce. validate_decision_envelope rejects every real allow. Every replay test passes because _request_from() rebuilds the request out of the binding, i.e. the enriched form: a self-consistent fake agreeing with itself, which hid this through three rounds of digest work. Third time a real artifact has beaten a fake in this integration. NOT FIXED, DELIBERATELY. Rejecting a valid allow is wrong in the safe direction. Which fields may be enriched is flex-auth's contract to publish; inferring it means accepting a binding that differs from our proposal in a way we decided was benign -- the fail-open shape GH-DEC-2026-008 rejected for vocabularies and FLEX-DEC-2026-007 for digests. Raised with them. Tenant question closed by operator decision 5ed3fb35: tenant:platform exactly, and service_auth.TENANT stays tenant:coulomb because the two identity layers are to remain distinct. Declining to author that mapping was right -- the answer was neither reading offered. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-07 00:22:12 +02:00
from secrets_engine.decision_check import (
check_decision,
require_supported_pdp_address,
)
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
from secrets_engine.errors import DecisionError
FIXTURES = Path(__file__).parent / "fixtures" / "flex-auth-replay"
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
PACKAGE_DIGEST = "sha256:bd11c5fe77ce6439c65fea225ad6b71d2110efc5e7b5bc9b499c59cd0a53b8b4"
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
SNAPSHOT_DIGEST = "sha256:f5a309bc0b36721fd6d9ad7f53eb21222162bc2eac62a0ab0802a9a1d51340bb"
CASES = {
"rotate": {
"file": "decision_rotate.json",
"digest": "sha256:de67324f54187055307a833235f83ced9fcd3a20952a27b3d19493ed39734345",
"action": "rotate",
"context_claim_digest": None,
},
"destroy": {
"file": "decision_destroy_dual_control.json",
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
"digest": "sha256:c749ee2dc3cdf927a70a3e5b27cff4d97a438d3264153b4b2e3bcacbaf82091a",
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
"action": "destroy",
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
"context_claim_digest": "sha256:8b73d29ecef286d42e03d2420531d6c45219f325a7ae004c1ecfc781203a2800",
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
},
}
def _envelope(name):
return json.loads((FIXTURES / CASES[name]["file"]).read_text())
def _request_from(envelope):
"""Rebuild the normalized tuple the binding carries.
Per the contract, a consumer re-hashing the *original unenriched* request
will not match a decision that turned on registry attributes; the binding is
the evaluator's statement of what it hashed.
"""
binding = envelope["binding"]
request = {"id": envelope["request_id"]}
for key in ("tenant", "subject", "action", "resource"):
if binding.get(key) is not None:
request[key] = binding[key]
if binding.get("context") is not None:
request["context"] = binding["context"]
return request
def _refresh_lifetime(envelope):
"""Lifetime moves with the clock and must never be pinned."""
now = datetime.now(timezone.utc)
envelope["lifetime"]["not_before"] = (now - timedelta(minutes=1)).strftime(
"%Y-%m-%dT%H:%M:%SZ"
)
envelope["lifetime"]["expires_at"] = (now + timedelta(minutes=14)).strftime(
"%Y-%m-%dT%H:%M:%SZ"
)
return envelope
@pytest.mark.parametrize("name", list(CASES))
def test_request_digest_matches_the_issued_decision(name):
envelope = _envelope(name)
expected = CASES[name]["digest"]
assert envelope["binding"]["request_digest"] == expected, "fixture drifted"
assert request_digest(_request_from(envelope)) == expected
@pytest.mark.parametrize("name", list(CASES))
def test_correlation_fields_are_not_hashed(name):
"""id, policy_version and caring_context must not move the digest."""
request = _request_from(envelope := _envelope(name))
baseline = request_digest(request)
assert baseline == envelope["binding"]["request_digest"]
for field, value in (
("id", "check:some-other-correlation-id"),
("policy_version", "v99"),
("caring_context", {"anything": "here"}),
):
assert request_digest({**request, field: value}) == baseline, field
stripped = {k: v for k, v in request.items() if k != "id"}
assert request_digest(stripped) == baseline
@pytest.mark.parametrize("name", list(CASES))
def test_digest_material_is_exactly_the_published_tuple(name):
material = digest_material(_request_from(_envelope(name)))
assert set(material) <= {"tenant", "subject", "action", "resource", "context"}
assert "id" not in material
@pytest.mark.parametrize("name", list(CASES))
def test_real_envelope_validates_against_the_published_package(name):
envelope = _refresh_lifetime(_envelope(name))
result = validate_decision_envelope(
envelope,
_request_from(envelope),
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
accepted_policy_versions={"v2"},
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
)
assert result.action == CASES[name]["action"]
assert result.subject_id == "secrets-engine"
@pytest.mark.parametrize("name", list(CASES))
def test_provenance_digests_are_pinned(name):
provenance = _envelope(name)["provenance"]
assert provenance["policy_package_digest"] == PACKAGE_DIGEST
assert provenance["registry_snapshot_digest"] == SNAPSHOT_DIGEST
assert provenance["evaluator"] == "flex-auth/local"
assert provenance["mode"] == "standalone"
@pytest.mark.parametrize("name", list(CASES))
def test_input_claim_digest_is_present_only_with_a_context(name):
"""The reason two fixtures exist: this field is conditional."""
provenance = _envelope(name)["provenance"]
expected = CASES[name]["context_claim_digest"]
actual = (provenance.get("input_claim_digests") or {}).get("context")
assert actual == expected
@pytest.mark.parametrize("name", list(CASES))
def test_a_tampered_binding_field_breaks_the_digest(name):
envelope = _envelope(name)
request = _request_from(envelope)
request["action"] = "handoff"
assert request_digest(request) != envelope["binding"]["request_digest"]
def test_expired_real_envelope_fails_closed():
"""The shipped lifetime is 15m from allow_ttl and has long since passed."""
with pytest.raises(DecisionError, match="lifetime has expired"):
envelope = _envelope("rotate")
validate_decision_envelope(
envelope,
_request_from(envelope),
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
accepted_policy_versions={"v2"},
fix: exclude correlation fields from the flex-auth request digest Verified the digest join against flex-auth's T03 replay fixtures and found request_digest was hashing fields docs/canonical-request-digest.md excludes. The material is tenant, subject, action, resource, context only: id is correlation, policy_version lives in provenance, caring_context is hashed separately. This engine included all three when present. Because the join adopts the served request id, every real production request would have carried one, so the computed digest would have matched no issued decision and failed closed against every correct allow. Same unsatisfiable shape as the removed AUTHORITY constant. The old pinned constant was computed with the id inside the material, so it was wrong and its passing proved nothing. Replaced with fixture-driven tests over two real envelopes (vendored with provenance) plus a structural test that correlation fields do not move the digest. Both fixtures are needed: input_claim_digests.context appears only with a non-empty context. Also stops computing the native claim digest. The claim's binding.action and binding.target speak approval-engine's vocabulary while ours speaks the catalog's, and no mapping is published; flex-auth makes no cross-check and states the correspondence is ours via pdp_digest. A claim recording no pdp_digest now fails closed naming the missing mapping rather than comparing two different languages. That mapping is a prerequisite for destroy. 274 tests pass. Production still fails closed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01M65ovP3eiiPHubibvWs9mD Assistant: claude-code Assistant-Model: opus Assistant-Process: 393550@bnt-lap001 Assistant-Session: 4bb359f9-1f12-4410-9e76-079cf23c82e4
2026-09-06 14:17:38 +02:00
)
def test_embedded_claim_uses_approval_engine_vocabulary_not_ours():
"""The published vocabulary gap, asserted rather than described.
The dual-control request carries a complete approval-claim in
context.approval. Its binding speaks approval-engine's language while the
decision speaks ours, and no mapping between them is published -- which is
why this engine compares pdp_digest and refuses to derive a native digest
from its own vocabulary.
"""
envelope = _envelope("destroy")
claim = envelope["binding"]["context"]["approval"]
assert claim["kind"] == "approval-claim"
assert claim["issuer"] == "approval-engine"
assert claim["valid_now"] is True
assert claim["binding"]["action"] == "secrets.kv.destroy"
assert envelope["binding"]["action"] == "destroy"
assert claim["binding"]["action"] != envelope["binding"]["action"]
assert set(claim["binding"]["target"]) == {"id", "stage"}
assert claim["binding"]["target"]["id"] != envelope["resource"]["id"]
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
APPROVAL_BINDING_DIGEST = (
"sha256:fa07becfaa471394d06aee5fa3cd66352bf0cc69ef24900240489684cda8cd56"
)
def test_embedded_claim_pdp_digest_cannot_equal_the_carrying_request_digest():
"""Carrying the claim inside a hashed context makes the two unequal.
context is part of the digest material, so embedding an approval-claim
changes the request digest of the very request that carries it. A
pdp_digest recorded at issue time therefore cannot equal the final digest
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
of the dual-control request. This is why comparing pdp_digest against
request_digest can never pass and would fail destroy closed forever.
"""
envelope = _envelope("destroy")
binding = envelope["binding"]
pdp = binding["context"]["approval"]["binding"]["pdp_digest"]
request = _request_from(envelope)
assert request_digest(request) == binding["request_digest"]
assert pdp != binding["request_digest"]
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
def test_pdp_digest_equals_the_published_approval_binding_digest():
"""FLEX-DEC-2026-007 closed the circularity, and we reproduce the value.
``approval_binding_digest`` is the canonical digest with context.approval
removed. Recomputing it here from our own canonical implementation is the
hermetic proof that this engine hashes the same material flex-auth does --
a pin alone would only assert the constant, not that we can derive it.
"""
envelope = _envelope("destroy")
binding = envelope["binding"]
pdp = binding["context"]["approval"]["binding"]["pdp_digest"]
assert binding["approval_binding_digest"] == APPROVAL_BINDING_DIGEST
assert pdp == APPROVAL_BINDING_DIGEST
assert approval_binding_digest(_request_from(envelope)) == APPROVAL_BINDING_DIGEST
def test_approval_binding_digest_is_not_a_replay_identity():
"""It must not collapse into request_digest, or an allow becomes replayable.
Two requests differing only in which approval was presented share an
approval_binding_digest while their decisions differ -- one allows, one
denies dual_control_required. The fixture asserts the two digests disagree
on a claim-bearing request so the distinction stays real.
"""
envelope = _envelope("destroy")
binding = envelope["binding"]
assert binding["approval_binding_digest"] != binding["request_digest"]
def test_ordinary_decision_carries_no_approval_binding_digest():
"""The field is omitted, not duplicated, on a claim-free decision."""
binding = _envelope("rotate")["binding"]
assert "approval" not in binding.get("context", {})
assert "approval_binding_digest" not in binding
def test_embedded_claim_declares_the_pdp_path():
"""pdp_path is the issuer's declaration, and it is what our PEP requires.
Path intent is never inferred from a pdp_digest that happens to be present;
approvals issued before approval-engine schema v3 carry pdp_path false
regardless of any digest they hold.
"""
claim = _envelope("destroy")["binding"]["context"]["approval"]
assert claim["binding"]["pdp_path"] is True
def test_decision_validation_ties_the_claim_to_the_approval_binding_digest():
"""Step 2 checks the identity against the real envelope, not a local guess.
The PEP knows the claim's pdp_digest from step 1. When the request it sent
carried that claim, the decision must name the same claim-free envelope, or
the approval was issued against some other request.
"""
envelope = _refresh_lifetime(_envelope("destroy"))
request = _request_from(envelope)
result = validate_decision_envelope(
envelope,
request,
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
accepted_policy_versions={"v2"},
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
expected_approval_binding_digest=APPROVAL_BINDING_DIGEST,
)
assert result.action == "destroy"
with pytest.raises(DecisionError, match="approval binding digest"):
validate_decision_envelope(
envelope,
request,
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
accepted_policy_versions={"v2"},
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
expected_approval_binding_digest="sha256:" + "c" * 64,
)
def test_claim_bearing_request_without_a_binding_digest_fails_closed():
"""A decision that records no binding digest cannot tie the claim to itself.
Comparing against request_digest instead would be the fail-open direction
the whole field exists to prevent, so the absence is refused outright.
"""
envelope = _refresh_lifetime(_envelope("destroy"))
request = _request_from(envelope)
del envelope["binding"]["approval_binding_digest"]
with pytest.raises(DecisionError, match="records no approval_binding_digest"):
validate_decision_envelope(
envelope,
request,
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
accepted_policy_versions={"v2"},
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
expected_approval_binding_digest=APPROVAL_BINDING_DIGEST,
)
def test_a_forged_binding_digest_is_recomputed_not_trusted():
"""The field is verified against our own canonical digest, never taken on faith."""
envelope = _refresh_lifetime(_envelope("destroy"))
request = _request_from(envelope)
envelope["binding"]["approval_binding_digest"] = "sha256:" + "d" * 64
with pytest.raises(DecisionError, match="does not match this request"):
validate_decision_envelope(
envelope,
request,
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
fix: send the tenant the policy package scopes us to, and adopt v2 build_action_request emitted no tenant field at all. The deployed secrets-engine.catalog-lane.lifecycle v2 package reads known_tenant := "tenant:platform" request_tenant := object.get(input, "tenant", "") so an absent tenant is not an ignored field, it matches the wrong_tenant first-denial branch. Every gated action this engine sent would have been denied -- and the omission also produced a request_digest that could match no correctly issued decision, since tenant is hashed material. That is the same class of defect as hashing excluded fields, arriving from the other direction, and again only a real artifact exposed it. Found by answering the GLAS-WP-0015 tenant-alignment question instead of assuming the values lined up. - REQUEST_TENANT is pinned against the vendored allow envelopes, so a package retenanting fails a test rather than denying production. - An empty tenant is refused at build time. - Accepted policy version moves v1 -> v2. v1 had no tenant rule and failed open: a rotate under tenant:coulomb returned allow against the deployed package. flex-auth superseded rather than amended it, because a fail-open correction has to be visible as a version change. A test pins that a v1 decision is refused. - Vendored decision_wrong_tenant_deny.json as the denial evidence glas asked for, with tests that we refuse it on effect before anything else and that a deny legally carries no lifetime. docs/tenant-alignment.md states the three tenant values as this repo holds them. It does not resolve the JWT/store mapping: service_auth.TENANT is tenant:coulomb, which is exactly the value the package denies. That is either two layers sharing a namespace format or one wrong constant, and picking between them without an owner ruling is the fail-open shape GH-DEC-2026-008 rejected for action vocabularies. Both constants stay as they are, deliberately not unified. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 22:32:54 +02:00
accepted_policy_versions={"v2"},
)
# --- tenant scoping (v2) -----------------------------------------------------
#
# GLAS-WP-0015 asked for the exact CheckRequest tenant and wrong-tenant denial
# evidence. These are that evidence, taken from a real deny envelope rather than
# asserted, plus the pin that stops our request tenant drifting from the package
# it is scoped by.
WRONG_TENANT_FIXTURE = "decision_wrong_tenant_deny.json"
def test_our_request_tenant_is_the_package_known_tenant():
"""The tenant we send must be the one the package allows.
v2 reads request_tenant := object.get(input, "tenant", "") and allows only
known_tenant := "tenant:platform". Pinning our constant against the real
allow envelopes means a package retenanting shows up here rather than as a
wrong_tenant denial in production.
"""
for name in ("rotate", "destroy"):
assert _envelope(name)["binding"]["tenant"] == REQUEST_TENANT
def test_built_request_carries_the_tenant_and_hashes_it():
"""An absent tenant is a denial, not an ignored field, so it must be sent.
tenant is part of the digest material, so omitting it also produces a digest
that matches no correctly issued decision -- the same class of defect as
hashing excluded fields, arriving from the other direction.
"""
entry = SimpleNamespace(id="glas-primary", stage="prod")
request = build_action_request(
entry,
"rotate",
subject_id="secrets-engine",
subject_type="service",
purpose="rotate-exposed-key",
fields=["password"],
)
assert request["tenant"] == REQUEST_TENANT
assert "tenant" in digest_material(request)
untenanted = dict(request)
del untenanted["tenant"]
assert request_digest(untenanted) != request_digest(request)
def test_build_action_request_refuses_an_empty_tenant():
entry = SimpleNamespace(id="glas-primary", stage="prod")
with pytest.raises(DecisionError, match="requires a tenant"):
build_action_request(
entry,
"rotate",
subject_id="secrets-engine",
subject_type="service",
purpose="rotate-exposed-key",
tenant="",
)
def test_wrong_tenant_deny_envelope_fails_closed():
"""Wrong-tenant denial evidence, from a real v2 deny envelope.
flex-auth sent tenant:coulomb on an otherwise-valid rotate and the package
denied it with matched_rule wrong_tenant, so the tenant alone carried the
denial. Our consumer must refuse it on effect before anything else -- a deny
is not a decision we may act on, whatever else it validates.
"""
# Not lifetime-refreshed: a deny carries no lifetime at all, which is the
# property test_wrong_tenant_deny_carries_no_lifetime pins.
envelope = json.loads((FIXTURES / WRONG_TENANT_FIXTURE).read_text())
assert envelope["effect"] == "deny"
assert envelope["reason"] == "wrong_tenant"
assert envelope["matched_rule"] == "wrong_tenant"
assert envelope["binding"]["tenant"] == "tenant:coulomb"
assert envelope["binding"]["tenant"] != REQUEST_TENANT
with pytest.raises(DecisionError, match="effect is not allow"):
validate_decision_envelope(
envelope,
_request_from(envelope),
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
accepted_policy_versions={"v2"},
)
def test_wrong_tenant_deny_carries_no_lifetime():
"""A deny has no lifetime, which is why effect must be checked first.
DecisionEnvelope requires lifetime only when effect is allow. A consumer
that validated lifetime before effect would raise a confusing missing-field
error on a perfectly well-formed denial.
"""
envelope = json.loads((FIXTURES / WRONG_TENANT_FIXTURE).read_text())
assert envelope.get("lifetime") is None
def test_the_superseded_v1_package_is_not_accepted():
"""v1 had no tenant rule and failed open; pinning it must not be possible.
A consumer still pinned to v1 would keep getting allows it should never
have had and could not tell from the version string that the rule changed
underneath it, which is exactly why flex-auth superseded v1 rather than
amending it.
"""
envelope = _refresh_lifetime(_envelope("rotate"))
with pytest.raises(DecisionError, match="policy version is not accepted"):
validate_decision_envelope(
envelope,
_request_from(envelope),
accepted_policy_packages={"secrets-engine.catalog-lane.lifecycle"},
feat: bind the destroy gate to approval_binding_digest and pdp_path The vocabulary mapping this path was waiting on is not coming: gate-house rejected it in GH-DEC-2026-008, because a translation can be confidently wrong and fails open by accepting a claim approved for a different action. The stronger option arrived instead, and both halves are enforced here. flex-auth published binding.approval_binding_digest (FLEX-DEC-2026-007) to fix the circularity this repo reported: a pdp_digest recorded at issue time can never equal the request_digest of the request that carries the claim in its hashed context, so with GH-DEC-2026-008 requiring that equality, destroy would have failed closed forever on a check no correct record could pass. - authorization.approval_binding_digest implements the published exclusion rule, including Go's context,omitempty behaviour when stripping empties the context; digest_material drops an empty context for the same reason. - validate_decision_envelope recomputes the field rather than trusting it, refuses a claim-bearing request whose decision records none, and compares the claim's digest from step 1 against it -- never against request_digest, which still covers the claim so it stays a sound replay identity. - validate_approval_claim requires binding.pdp_path true before using pdp_digest at all. Path intent is never inferred from a digest that happens to be present; pre-schema-v3 approvals carry pdp_path false regardless of any digest they hold. Replay fixtures re-vendored from dd3ce4c. The destroy pins moved a second and final time; approval_binding_digest did not, which is the point. The fixture now demonstrates the property instead of asserting it: we rederive fa07becf... from its own request through our canonical implementation, proving we hash the same material flex-auth does rather than pinning a constant we cannot reproduce. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-06 20:39:59 +02:00
accepted_policy_versions={"v1"},
)
feat: adopt the owner-documented PDP access path, and prove it live Adoption asked for by flex-auth (FLEX-WP-0021-T05) and glas-harness (GLAS-WP-0015), plus the first real decision this engine has obtained from the deployed pin -- which found a defect the fixtures could not. ACCESS PATH. require_supported_pdp_address refuses in-cluster Service names and any non-loopback host. This is no longer a unilateral call: the owner path is documented as loopback kubectl port-forward over the authenticated Kubernetes API, which is what authenticates the responder transitively (FLEX-DEC-2026-010). A Service name from a workstation does not fail, it resolves through the DNS search suffix to an unrelated public host, and since decision records carry no signature, a responder knowing the published package and version can return an allow that passes every check we make. Fail-closed protects against a PDP that is absent, not one that lies. The guard runs before the token is read, so a misdirected request cannot leak it; a test pins that ordering. LIVE PROOF. Minted a 10-minute TokenRequest token (audience flex-auth, SA secrets-engine/secrets-engine, mode 0600 outside the worktree, shredded after), forwarded to the named pod, and sent a real CheckRequest for glas-claude-agent-dev-anthropic. Result: allow, catalog_lane_policy_matched, served by v2 (sha256:bd11c5fe...) -- so the redeploy flex-auth flagged as outstanding has landed and the pin no longer serves the tenant-blind v1. Our tenant fix is confirmed against the real service: binding.tenant is tenant:platform. THE DEFECT IT FOUND. The evaluator enriches from its registry before hashing -- subject gains attributes and tenant, resource gains tenant -- so binding.request_digest is over material we never sent and cannot reproduce. validate_decision_envelope rejects every real allow. Every replay test passes because _request_from() rebuilds the request out of the binding, i.e. the enriched form: a self-consistent fake agreeing with itself, which hid this through three rounds of digest work. Third time a real artifact has beaten a fake in this integration. NOT FIXED, DELIBERATELY. Rejecting a valid allow is wrong in the safe direction. Which fields may be enriched is flex-auth's contract to publish; inferring it means accepting a binding that differs from our proposal in a way we decided was benign -- the fail-open shape GH-DEC-2026-008 rejected for vocabularies and FLEX-DEC-2026-007 for digests. Raised with them. Tenant question closed by operator decision 5ed3fb35: tenant:platform exactly, and service_auth.TENANT stays tenant:coulomb because the two identity layers are to remain distinct. Declining to author that mapping was right -- the answer was neither reading offered. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E4tNMAYcSQmZWUE4wqP4ij Assistant: claude-code Assistant-Model: opus Assistant-Process: 715726@bnt-lap001 Assistant-Session: 80a42b32-cba6-4b23-8be0-68819b1a6092
2026-09-07 00:22:12 +02:00
# --- supported PDP address (FLEX-DEC-2026-010) -------------------------------
@pytest.mark.parametrize(
"url,fragment",
[
# The exact address flex-auth originally handed over. It does not fail
# to resolve from a workstation, it resolves to an unrelated public host.
("http://flex-auth-secrets-engine.flex-auth.svc.cluster.local:8080",
"in-cluster Service name"),
("http://flex-auth-secrets-engine.flex-auth.svc.cluster.local.:8080",
"in-cluster Service name"),
("http://flex-auth-secrets-engine.flex-auth.svc:8080",
"in-cluster Service name"),
# A public address is refused even though it would "work": nothing
# authenticates the responder, so reaching something is not reaching
# the pin.
("http://80.158.43.29:8080", "is not loopback"),
("https://flex-auth.example.com", "is not loopback"),
],
)
def test_unsupported_pdp_addresses_are_refused(url, fragment):
with pytest.raises(DecisionError, match=fragment):
require_supported_pdp_address(url)
@pytest.mark.parametrize(
"url",
[
"http://127.0.0.1:18080",
"http://localhost:18080",
"http://[::1]:18080",
],
)
def test_loopback_forward_addresses_are_accepted(url):
require_supported_pdp_address(url)
def test_the_guard_runs_before_the_token_is_read(tmp_path):
"""A bad address must not cause the bearer token to be read, let alone sent.
The token is the thing a misdirected request would leak, so the address
check has to come first rather than alongside.
"""
missing = tmp_path / "never-read.token"
with pytest.raises(DecisionError, match="in-cluster Service name"):
check_decision(
base_url="http://flex-auth-secrets-engine.flex-auth.svc.cluster.local:8080",
token_file=missing,
request={},
)
assert not missing.exists()