| Security meaning and conformance | security-canon, net-kingdom, gate-house, info-tech-canon | Versioned source contracts, offline validators and published policy references; no mandatory runtime hop for document interpretation. |
| Authentication and assurance | key-cape; user-engine for account journeys | Existing identity stack and explicit relying-party adapters. Account, group and factor recovery must work end to end before considering an identity-provider replacement. |
| Tenant and other facts | tenant-engine and the relevant fact owner | Use only policy-required facts with explicit provenance and freshness. Zone and maturity reference tools are not implicitly live claim services. |
| Authorization | access-engine | One decision owner and contract, with independently admitted consumer pins. Shared deployment requires a measured isolation and availability case. |
| Approval and presentation | approval-engine; informed-decision | One bound approval lifecycle, separate presentation; no approval inferred from a report, case type or scheduler. |
| Credential lifecycle and custody | secrets-engine; railiance-platform/OpenBao | Native identity first, scoped short-lived credentials next, managed static credentials only where necessary. Secrets Engine remains a CLI/workflow capability. |
| Operational access and construction | ops-warden, ops-mason, ops-bridge | Reuse routing, case follow-through, reviewed structure and private transport; no general umbrella broker. |
| Independent evidence | audit-core | Durable custody and admitted reconciliation; original owner receipts stay authoritative. State Hub indexes references rather than becoming security custody. |
| Agent execution | glas-harness, sand-boxer, the selected rein | An admitted task identity and enforced profile; tools and inherited credentials count as effective authority. |
| Deployment and recovery | existing Railiance, rail, reef and rapp owners | Pinned packages, owner deployment paths, off-host recovery resources, measured headroom and declared rollback. |
| Discovery and improvement | reuse-surface, repo-manager, kings-guard, maturity-engine, activity-core, coordination-engine | Derived discovery and observations feed deduplicated owner work. Scheduling and scores never create permissions. |
## Consumer profiles
### Application access
An application authenticates the caller using its admitted issuer, verifies the
required identity/tenant/assurance claims, obtains an Access Engine decision, and
enforces the exact authorized resource/action. User or tenant lookup is included
only when required by that application's policy. Cached facts need an explicit
freshness and revocation contract. Policy-required unavailable or stale facts
| TSA-03 | Crashes preserve uncertainty and cleanup | Actual ledger restart/concurrency, consumed-before-effect failure, target reconciliation, scoped cleanup and revocation evidence; no blind replay. |
| TSA-04 | First reusable browser case is accepted | NK-WP-0048 and named Mason/Warden/KeyCape adapters; independent native structure, sign-in/denial, delivery, revocation and rollback. |
| TSA-05 | Human and browser consumers can use the delivered identity | KEY-WP-0040/U08 real-factor recovery; User Engine journey; NK-WP-0050 and CUST-WP-0080-T05 human/agent Playwright acceptance. |
| TSA-06 | Actual agent and runner authority is bounded | CUST-WP-0073/GLAS-WP-0012 admitted runtime read-denial/tool-effect proof; Forge/Platform per-job credential and repository-workflow abuse tests. |
| TSA-07 | Credential lifecycle is inventoried and simplified | Current owner/consumer denominator, unknown/orphan findings, native-identity elimination candidates; sampled replacement/adoption/old-refusal and deletion semantics. G04/G09/G12. |
| TSA-08 | Recovery is independent and measured | NK-WP-0049 plus platform/root owners; isolated recovery with ordinary issuer unavailable, restore/reconnect/audit continuity and measured RTO/RPO. G08/G11. |
| TSA-11 | Discovery remains accurate | All reviewed IDs exactly once, owner/index parity, usable source links, current limits/version aliases, drift checks and package-family disposition. Reuse/Repo Manager/provider owners. |
| TSA-12 | Improvement operates without repeated manual polling | CUST-WP-0079/Activity accepted schedule, unassisted firing, checkpoint and changed-state notifications; adopted cadence runtime/receiver receipts; owner change and native remeasurement. |
| TSA-13 | Shared adapters reduce per-consumer implementation | Producer-format cross-language fixtures, two real consumers, version compatibility, denied-path parity, upgrades and rollback; no second PDP. |
| TSA-14 | Minimal profiles are independently reusable | Version/provider lock, dependency and optional-provider declaration, provider adapters, second clean installation, complete positive/negative journeys and upgrade/recovery instructions. |
| TSA-15 | Optimization improves measured operations | Baseline and repeat measurements of onboarding time/touches, p50/p95 decision/effect latency, scanner throughput/failure, resources/headroom and owner effort with unchanged control coverage. |
| TSA-16 | Governance and exceptions remain truthful | Existing SEC-DSM ledger and close-down controls reconciled to exact evidence; real authority for changes, explicit residual promotions, preserved deferrals and no invented owner assent. |
## Rollout and exit
Start authenticated decisions and the bounded metadata operation, with audit
acceptance included. Agent isolation is a parallel priority. Finish the existing
browser case rather than opening a third reference journey. Inventory, recovery
preparation, discovery corrections and source checks may proceed alongside these
two integrations. Package accepted interfaces and extend consumer coverage next.
Changes use owner repositories and their existing workplans. Scoped independent
owner reviews precede required native activation. The standing build-mode rule
already authorizes qualifying non-destructive changes with reviewed diff/dry-run,
rollback, abort owner and receipts. Access/secret policy scope changes, destructive
operations, new spend ceilings and public publication retain their explicit gates.
Existing exact approvals are reused within scope, not requested again.
Rollback either restores the previously admitted behavior within its accepted
exception or leaves the affected operation unavailable. It never silently accepts
an unsigned decision, widens scope or resurrects a revoked credential. Native
failure remains evidence and prevents further promotion until resolved.
The goal completes only when TSA-01 through TSA-16 have scope-matched acceptance
evidence or an already applicable explicit governing exception, the second
consumer/installation proves reuse, and no required work is left only in prose.
Evidence review must inspect actual sources and outcomes; a green aggregate