‹ Mixoff — Office Productivity SuiteQuality report QR-26-00102
Mixoff — Office Productivity Suite
Local-first Windows productivity suite with Documents, Sheets, Slides, Mixoff Flow, and optional AI integration.
- Product ID
- B0814-15415
- Builder
- Zyphase
- Version reviewed
- 1.0.0 · released 6 Nov 2026
- Stack
- React 18.3.1 with Tauri 2 (@tauri-apps/api 2.11.1) · TypeScript 5.9.3 (frontend) with Rust 1.99.1 (native layer) · None; optional managed-AI service uses SQLite via node: 24.22.1
- Built with
- OpenAI Codex
- Review period
- 30 Oct 2026 – 6 Nov 2026 · 2 rounds of changes
- ✓AI code audit6 checks · 29 Oct 2026
- ✓Installed & run6 Nov 2026
- ✓Demo vs delivered6 Nov 2026
- ✓Final sign-off6 Nov 2026
At a glance
- ✓Installed and run from the delivered code — 8 of 8 checks answered yes.
- ✓Live demo compared with the delivered code — 7 of 7 checks answered yes.
- !16 of 21 promised items delivered (setup guide, example config and the rest).
- !Security scored 74/100 with no failing check.
- !5 open warnings — explained in part 1.
AI code audit
The full source archive was read by our AI auditor against each check. It reads the code — it does not run it; running it is part 2.
| Check | What we found | Score | Result |
|---|---|---|---|
| Secrets & credentials | Every regex candidate was reviewed and none contains a real credential. The managed-ai .env.example uses clear REPLACE_WITH placeholders, and the other hits are secret-detection regexes, runtime handling of user-supplied credentials in test mocks, and tokenizer function names. Only minor hygiene notes remain. | 96 | Pass |
| Dependency vulnerabilities | Lockfiles exist for npm (package-lock.json) and Rust (Cargo.lock) and direct dependencies are exact-pinned, which is good. However, the xlsx dependency is installed from a SheetJS CDN tarball outside the npm registry, and the managed-ai service has no lockfile or declared dependencies. Several pinned versions (e.g. vitest 5.0.5, vite 8.3.4, @tiptap 3.31.3) could not be verified against known advisories from the digest, so the lockfile contents were not audited line by line. | 72 | Warning |
| Authorisation & access control | The product is a local-first Tauri desktop app, so its main attack surface is a small managed-AI gateway whose routes (gateway.mjs, server.mjs) were not included in the excerpts. The README describes bearer device tokens for customer routes and a separate admin secret for /admin/*, but I could not verify the guards, ownership checks or constant-time comparison in code. The demo is a static mock-IPC frontend with no data-changing backend. | 55 | Warning |
| Database access rules | The desktop app stores data in native files, not a database, so most SQL checks do not apply. The one database is the managed-ai gateway's SQLite ledger (gateway.mjs, server.mjs). The digest did not include its source, so injection, unscoped UPDATE/DELETE, per-user scoping, foreign keys and indexes could not be verified. The README also admits the ledger is not production-ready for multi-host use. | 70 | Warning |
| Code structure & duplication | The digest shows a sensible top-level split (src/modules, src/flow, src/storage, src-tauri, crates/file-safety), but several files are written as dense one-line code that is hard to review. The scripts folder and test harnesses also appear to contain large amounts of generated or duplicated tooling that buyers would have to maintain. | 62 | Warning |
| Build reproducibility | The main app has pinned dependencies, lockfiles, and documented npm/Tauri build steps, so a Windows machine can likely restore and build it. However, the archive is nested with a duplicated top-level a project file, README links point to files outside the digest, and several prerequisites (Python, Edge, Rust version, Node for the gateway) are not clearly pinned. Setup in HowToSetup.md could not be verified from the digest. | 72 | Warning |
Open warnings
Problems that did not stop the listing. They stay on the product page until a later version fixes them.
The digest shows a sensible top-level split (src/modules, src/flow, src/storage, src-tauri, crates/file-safety), but several files are written as dense one-line code that is hard to review. The scripts folder and test harnesses also appear to contain large amounts of generated or duplicated tooling that buyers would have to maintain.
The desktop app stores data in native files, not a database, so most SQL checks do not apply. The one database is the managed-ai gateway's SQLite ledger (gateway.mjs, server.mjs). The digest did not include its source, so injection, unscoped UPDATE/DELETE, per-user scoping, foreign keys and indexes could not be verified. The README also admits the ledger is not production-ready for multi-host use.
The product is a local-first Tauri desktop app, so its main attack surface is a small managed-AI gateway whose routes (gateway.mjs, server.mjs) were not included in the excerpts. The README describes bearer device tokens for customer routes and a separate admin secret for /admin/*, but I could not verify the guards, ownership checks or constant-time comparison in code. The demo is a static mock-IPC frontend with no data-changing backend.
Lockfiles exist for npm (package-lock.json) and Rust (Cargo.lock) and direct dependencies are exact-pinned, which is good. However, the xlsx dependency is installed from a SheetJS CDN tarball outside the npm registry, and the managed-ai service has no lockfile or declared dependencies. Several pinned versions (e.g. vitest 5.0.5, vite 8.3.4, @tiptap 3.31.3) could not be verified against known advisories from the digest, so the lockfile contents were not audited line by line.
The main app has pinned dependencies, lockfiles, and documented npm/Tauri build steps, so a Windows machine can likely restore and build it. However, the archive is nested with a duplicated top-level a project file, README links point to files outside the digest, and several prerequisites (Python, Edge, Rust version, Node for the gateway) are not clearly pinned. Setup in HowToSetup.md could not be verified from the digest.
Requirement list
What the builder agreed to deliver before uploading — 16 of 21 met.
- Complete source for the desktop app Met
- Lockfiles and pinned dependencies Partial
- No real credentials Met
- HowToSetup.md for each desktop platform Partial
- README.md with the five differentiating features Partial
- Reachable demo Partial
- Example settings file if the app reads keys (nice)Met
- CLAUDE.md for AI coding tools (nice)Met
- CHANGELOG.md starting at 1.0.0 (nice)Partial
- Tests for core engines (nice)Met
- Architecture documentation (optional)Met
- README.md Met
- HowToSetup.md Met
- Mixoff a project file Met
- Mixoff a project file Met
- Mixoff a project file Met
- Mixoff a project file Met
- Mixoff a project filee (nice)Met
- CLAUDE.md (nice)Met
- CHANGELOG.md (nice)Met
- LICENSE-NOTE.md Met
Human review
Reviewers from the Vibe96 team in Dubai worked through each question below; a final reviewer signed it off.
Stage 1 — Installed and run from the delivered code
Completed 6 Nov 2026A reviewer installed the product on a clean machine, following only the guide that ships with it.
| Question | Answer | Reviewer's note |
|---|---|---|
| Packages restored without errors | Yes | |
| Solution builds clean | Yes | |
| Database created from the shipped script or migrations | Yes | |
| First sign-in works with the documented steps | Yes | |
| Setup guide is accurate and complete | Yes | |
| Every module in the feature list opens and works | Yes | |
| No real secrets in the archive | Yes | |
| Branding and company details can be changed without code | Yes |
Stage 2 — Live demo compared with the delivered code
Completed 6 Nov 2026A second reviewer checked that what the listing and demo show is what the buyer receives.
| Question | Answer | Reviewer's note |
|---|---|---|
| Demo URL is reachable | Yes | |
| Demo credentials work | Yes | |
| Demo runs the same version as the archive | Yes | |
| All five differentiators are visible in the demo | Yes | |
| Screenshots match the demo | Yes | |
| Nothing in the demo is missing from the code | Yes | |
| Datasheet claims are accurate | Yes |
Changes asked for during review
- None — approved on the first round.
- Sign-off: approved for sale on 6 Nov 2026 with 5 warnings shown publicly.
What this report covers — and what it doesn't
- It covers version 1.0.0 exactly as the buyer downloads it: the source archive, the setup guide and the live demo on the dates above.
- It is not a penetration test, a performance or load test, or a legal review of third-party package licences.
- Scores are a point in time. Libraries age; a later version is audited again and gets its own report.
- Your setup matters. Hosting, configuration and changes made after purchase are outside this review.
- How the score is made: each AI check is scored 0–100 and weighted (security counts most). A listing needs 75 or more, no failing check, and the human review passed.
- Always try the live demo before buying, and ask the builder questions from the product page.
Vibe96 FZ-LLC, Dubai, UAE. This report describes the product at the time of review and is not a warranty.