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>

specs/001-persona-engine/contracts/cli-invoke-persona-engine.md

@@ -0,0 +1,90 @@
+# Contract: `Invoke-PersonaEngine.ps1`
+
+The engine entry point. Retrieval, evaluation, reporting, and controlled persistence.
+
+## Signature
+
+```powershell
+[CmdletBinding(SupportsShouldProcess = $true, ConfirmImpact = 'High')]
+param(
+    [Parameter(Mandatory)][string] $ConfigPath,
+    [Parameter()][guid]   $UserObjectId,
+    [Parameter()][string] $OutputPath,
+    [Parameter()][guid]   $CorrelationId
+)
+```
+
+`SupportsShouldProcess` supplies `-WhatIf` and `-Confirm`. `-Verbose` and `-Debug` are common
+parameters and are **not** declared.
+
+## Parameter contract
+
+| Parameter | Required | Behaviour |
+| --- | --- | --- |
+| `-ConfigPath` | Yes | Path to the JSON configuration. Validated through all four layers before any connection is made (FR-002). |
+| `-WhatIf` | No | **The approved no-write control.** Reads, evaluation, comparison, console output, summaries, and audit records all behave identically to enforcement; zero write requests are issued (FR-017, SC-004). |
+| `-UserObjectId` | No | Single-user execution for validation. Skips enumeration; retrieves one user. |
+| `-OutputPath` | No | Overrides the configured audit output path where permitted. |
+| `-CorrelationId` | No | Supplied run identifier. Generated when absent. Appears on every audit record. |
+| `-Verbose` | No | Operational detail. **MUST NOT** alter write behaviour. |
+| `-Debug` | No | Enables condition-value tracing (Principle V). **MUST NOT** imply read-only — a `-Debug` run without `-WhatIf` writes. |
+
+## Mode determination
+
+```text
+$PSCmdlet.ShouldProcess() returns $false  ->  Preview mode  ->  no write request constructed or sent
+$PSCmdlet.ShouldProcess() returns $true   ->  Enforce mode  ->  write permitted, subject to FR-016
+```
+
+Mode MUST be derived from `ShouldProcess` alone. A separate boolean "preview" flag is prohibited —
+two sources of truth for the write gate is precisely the defect class Principle III exists to
+prevent.
+
+## Write gate (FR-016)
+
+A write is issued only when **all** hold:
+
+1. Evaluation completed successfully (`Outcome != EvaluationError`).
+2. `CalculatedPersona != StoredPersona` (ordinal comparison, case-sensitive for change detection).
+3. The target attribute is non-blank and present in `approvedWritableAttributes`.
+4. `ShouldProcess` returned `$true` for this user.
+
+Failing any of these yields `Unchanged`, `WouldUpdate`, or `Skipped` — never a silent write.
+
+## Output contract
+
+- **Per user, immediately after evaluation** (FR-018, SC-012): one console line carrying UPN,
+  Account Object ID, outcome, matched rule ID, stored value, calculated value, and action.
+- **Every `summaryInterval` users** (FR-019): a table of all business rules with match counts, plus
+  outcome totals and a reconciliation check.
+- **At completion**: a final summary regardless of interval, including when the interval is `0`
+  (FR-020).
+- **Reconciliation** at every summary: `Processed = Matched + Unclassified + EvaluationError`
+  (FR-021). A mismatch is logged as an engine defect, at `Error` severity.
+- **Audit records**: see [audit-record.md](audit-record.md).
+
+## Exit codes
+
+| Code | Condition |
+| --- | --- |
+| `0` | Successful run; no fatal processing errors |
+| `1` | Configuration validation failure |
+| `2` | Authentication / authorization failure |
+| `3` | User enumeration failure |
+| `4` | Fatal required data-provider failure |
+| `5` | Reconciliation failure |
+| `6` | Unexpected fatal engine error |
+
+Every code MUST be reachable and returned for its documented condition (SC-011). A per-user
+`EvaluationError` does **not** by itself terminate the run; the final status reports the affected
+count and applies `evaluationErrorThreshold` when configured.
+
+## Invariants (test-asserted)
+
+| Invariant | Assertion |
+| --- | --- |
+| Zero writes under `-WhatIf` | The write adapter is mocked; call count is `0` over a full synthetic population (SC-004) |
+| Single-attribute body | Every captured request body has exactly one key, equal to `engine.targetAttribute` (SC-005) |
+| Idempotence | Second consecutive run over unchanged input issues zero writes (SC-002) |
+| Determinism | Same fixture set, shuffled input order, identical results (SC-003) |
+| Exactly one outcome | Every processed user appears in exactly one outcome bucket (SC-001) |