# Verification for Ember 1.0

Verified on October 9, 2026.

## Completed

- Compiled the Windows x64 GUI executable using the .NET Framework compiler and WPF assemblies; no build-time package downloads.
- Validated 71 unique catalog entries: 40 registry settings, 19 optional app removals and 12 guides. Verified unique registry targets, required benefit/tradeoff text, and no app removal in the recommended profile.
- Exercised the production journal/apply/restore logic against an in-memory registry: original values and types survive XML serialization, including unsigned DWORD bit patterns, QWORD extremes, unexpanded environment strings, binary data, and multi-strings.
- Verified restoring originally absent values, preserving conflicting newer values, retrying conflicts, repeated restore, and partial write failures. Machine-wide paths and unlisted app packages are rejected.
- Verified UI profile selection, search matches, empty search results, category filtering, clearing selections, and unselected app-removal defaults. Constructed all five pages without applying changes.
- Rendered and visually inspected the native WPF overview, optimization list, debloat page, recovery page, and minimum-size overview (1020 × 720, with scrolling).
- Ran the real read-only scanner outside the restricted development sandbox: it detected the Ryzen 5 7600X, integrated Radeon graphics, 32 GB RAM at a firmware-reported 6000 MT/s, Samsung 990 EVO SSD, Windows version, motherboard and graphics driver without scan warnings.
- Ran real read-only setting/app inspection: all 71 entries returned a status, distinguishing matching values, differing values, installed apps, absent apps and guides.

## Limits

The separate setup executable also passed isolated install/uninstall tests: embedded payload extraction, hash verification, existing-folder protection, clean removal of managed files, preservation of modified files, and rejection of traversal paths. These tests wrote only within a scratch folder and did not create real shortcuts or an Installed apps registry entry. Shortcut creation and Windows app-registration integration were reviewed in source, not exercised on the user's live profile.

No optimization was applied and no app was uninstalled from the user's PC during development. Registry mutation/recovery was tested with an isolated in-memory store and real disk journals, not by changing the user's gaming settings. Actual Appx removal was not executed as a test. This build is unsigned and has not been certified or tested on every Windows release or policy configuration.

No game benchmark or input-latency measurement was performed. There is no measured FPS improvement claim. Registry state matching is not proof that a Windows build honors a preference.

The scanner initially received access-denied errors inside the development sandbox. Normal read-only execution outside that sandbox successfully returned the hardware. The app exposes unavailable values and includes scan warnings in its exported report when Windows blocks access.

## Reproduce the isolated checks

Build from `source/build.ps1`, then use the following executable modes with writable absolute output paths. The app runs silently and writes results to the specified JSON files; wait for it to exit before opening those files.

```text
Ember.exe --self-test C:\YourFolder\scratch C:\YourFolder\engine-results.json
Ember.exe --ui-test C:\YourFolder\ui-results.json
Ember.exe --scan C:\YourFolder\hardware.json
Ember.exe --inspect C:\YourFolder\states.json
Ember.exe --preview C:\YourFolder\overview.png Overview
```

The `--self-test` mode uses simulated values only. `--ui-test` and `--preview` use clearly labeled user-supplied hardware for design verification and block apply operations. Normal launch automatically scans the real PC.