Implement Stage A: rule engine, validation, audit, and safety gates

Completes 109 of 121 tasks. Every remaining task needs a tenant connection
(T055, T056, T101-T103) or an Azure Automation account (T115-T121).

  354 offline Pester tests      PASS
  Engine purity (Principle IV)  PASS
  Sanitization (SC-013)         PASS  (156 files)
  Graph module loaded in tests  none  (SC-008 holds)

What landed
  - Four-layer configuration validation with stable finding codes, covering
    every VR-002 and VR-003 condition, plus a 23-fixture invalid-config corpus
  - Run loop, audit records (NDJSON through a single sink), summaries,
    reconciliation, and exit codes 0-6
  - Persistence behind a single write-body builder whose result always has
    exactly one key
  - Invoke-PersonaEngine.ps1 and Edit-PersonaEngineConfig.ps1
  - Six docs, two pipelines, traceability matrix, V-5a and sanitization records

Three deviations from tasks.md, each recorded in its status block

  T033 is not in Resolve-UserPersona. evaluationErrorThreshold is run-level
  state and the rule engine is pure; a counter there would break Principle IV.
  It lives in New-PersonaRunCounter and is applied in the run loop.

  A new src/Engine/ layer holds Invoke-PersonaEngineRun. The entry script
  imports the manifest, which requires Microsoft.Graph.Authentication, so a
  loop living only inside it could not run on a machine without the Graph SDK
  and SC-004 could not be proven at all. The entry script is now a thin
  wrapper and what ships is what is tested.

  The invalid-config corpus is generated by a committed script, with the
  generated fixtures committed too, so a reviewer sees the fixture in the diff.

Defects found by running the code, not by reading it

  Group and role ID lists were double-wrapped: @(Get-PersonaGroupIdPage ...)
  around a comma-returned array collapsed every membership list into one
  bogus space-joined entry. That is a silent false non-match, exactly what
  FR-013 exists to prevent.

  A 403 whose status appears only in the exception message parsed as $null,
  which the retry policy treats as a transport error - five requests per
  account against a tenant already refusing. Status extraction now falls back
  to the message text, bounded to 400-599.

  The sanitization scan walked tracked files only, so it covered 34 of 156
  files and none of this phase's code. It now scans untracked non-ignored
  files too, and a negative control confirms it catches a planted leak.

  Test-Json reports one error per violating location, not first-failure-only
  as the V-5a draft claimed. Record and pin corrected.

Enforcement remains blocked on the V-4 security sign-off.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This commit is contained in:
2026-08-20 21:48:19 -04:00
parent c59c85dd55
commit cdc6bb33d3
124 changed files with 16638 additions and 199 deletions
@@ -0,0 +1,124 @@
# Contract: Directory Data Provider
The only component permitted to talk to Microsoft Graph. Implements OTD-002, OTD-004, and OTD-007.
Everything below the normalization boundary is invisible to the rule engine (Principle IV).
**Transport**: `Invoke-MgGraphRequest` from `Microsoft.Graph.Authentication` (OTD-004). No
resource-specific SDK modules.
## Authentication (FR-003)
| Function | Environment | Mechanism |
| --- | --- | --- |
| `Connect-PersonaGraphManagedIdentity` | Azure Automation | `Connect-MgGraph -Identity` |
| `Connect-PersonaGraphInteractive` | Local development | `Connect-MgGraph -Scopes <read scopes>` |
Both return an opaque connection handle. No token, header, or secret is ever returned to a caller,
logged, or written to an audit record.
## Permissions (OTD-002)
| Function | Application permission |
| --- | --- |
| `Get-PersonaUsers` | `User.Read.All` (or `User.ReadWrite.All` for the enforcement identity) |
| `Get-PersonaGroupMembership` | `GroupMember.Read.All` |
| `Get-PersonaDirectoryRoles` | `RoleManagement.Read.Directory` |
| `Set-UserPersonaAttribute` | `User.ReadWrite.All` |
`Directory.Read.All` is prohibited — materially broader than the three read permissions combined.
## Read operations
### `Get-PersonaUsers`
```text
GET /v1.0/users?$select=<fields>&$top=999
```
- `$select` carries the FR-005 baseline (`id`, `userPrincipalName`, `displayName`, `userType`,
`accountEnabled`, `companyName`, `jobTitle`, `department`) plus the configured target attribute and
any property referenced by an enabled rule. Unused properties are not requested.
- The persona directory extension is selected by its full name,
`extension_<EXTENSION-APP-ID>_<APPROVED-PERSONA-ATTRIBUTE-NAME>`.
- Pagination follows `@odata.nextLink` until absent (FR-004). A truncated enumeration MUST raise —
never return a partial population as if complete.
- `-UserObjectId` switches to `GET /v1.0/users/{id}?$select=...`.
### `Get-PersonaGroupMembership`
```text
GET /v1.0/users/{id}/memberOf # direct
GET /v1.0/users/{id}/transitiveMemberOf # transitive
```
Mode comes from the condition, falling back to `engine.defaultMembershipMode` (RE-007).
**Return contract**: always a `MembershipRecord`. On failure it returns a record with
`RetrievalSucceeded = $false` and a `FailureReason` — it MUST NOT return an empty list, and MUST NOT
throw past the per-user boundary. This single behaviour is what makes FR-013 work: unknown membership
becomes `EvaluationError`, never a false non-match.
### `Get-PersonaDirectoryRoles`
```text
GET /v1.0/roleManagement/directory/roleAssignments?$filter=principalId eq '<id>'
```
Eligible (PIM) assignments are out of scope for v1 unless authorization is confirmed.
### Caching
Group and role data reusable across users is cached for the run's lifetime (NFR-002). The cache is
keyed by group or role Object ID and is **never** persisted between runs — a stale cache would make
results depend on run history, breaking Principle I.
## Write operation
### `New-PersonaWriteBody`
The **only** function permitted to construct a write body (OTD-003 control 3).
```powershell
# Returns exactly one key.
@{ "<engine.targetAttribute>" = "<CalculatedPersona>" }
```
Contract:
- Throws if the attribute name is not `engine.targetAttribute`.
- Throws if the attribute is absent from `approvedWritableAttributes`.
- Returns a hashtable whose `Count` is exactly `1`. Tests assert on this directly (SC-005).
### `Set-UserPersonaAttribute`
```text
PATCH /v1.0/users/{id}
Content-Type: application/json
<body from New-PersonaWriteBody>
```
- Callable **only** when `ShouldProcess` returned `$true`. Under `-WhatIf` this function is not
reached — the caller does not construct a request at all (SC-004). Preview mode is an absence of a
call, not a suppressed call.
- A failure returns `UpdateFailed` for that user and does not terminate the run.
## Retry policy (OTD-007)
| Aspect | Value |
| --- | --- |
| Retryable | 429, 500, 502, 503, 504, transport timeout |
| Never retried | 400, 401, 403, 404, 409 |
| `Retry-After` | Honoured when present; overrides computed backoff |
| Attempts | Max 5 |
| Backoff | Exponential from 1s, full jitter, per-delay cap 60s |
| Logging | Attempt number, status code, and delay on every retry |
Exhausted retries on **required** data produce `EvaluationError` for the affected user (FR-013).
Exhausted retries during enumeration are fatal (exit code `3`).
## Prohibited in this layer
- Logging tokens, `Authorization` headers, or full response bodies (Principle V).
- Returning raw Graph objects past `ConvertTo-Persona*Record`.
- Any reference to rule, persona, or condition concepts — this layer moves data, it does not decide.