Skip to content

Releases

v2026.10.1 — Expanded authorized security-assessment coverage with GraphQL operation-level checks, a streamlined New Assessment launcher, and a rebuilt Windows release pipeline (Electron 44.7.0).

Choose a tag to compare

@vvts-alpha vvts-alpha released this 19 Oct 04:31

Summary

Expanded authorized security-assessment coverage with GraphQL operation-level checks, a streamlined New Assessment launcher, and a rebuilt Windows release pipeline (Electron 44.7.0).

Changes

  • Pilot probe catalog expanded: first-class GraphQL operation inventory and per-operation authorization/IDOR checks, WebSocket, deserialization, XXE, cache, method-bypass, rate-limit, OAuth/OIDC, SAML, request-smuggling, CRLF, and host-header probes.
  • Authentication evidence is now role/session-specific: whoami liveness is tracked separately from authenticated-content proof, and role-specific crawl frontier/visited state survives role switches and SQLite resume.
  • New Assessment launcher reorganized: compact Run/Target/launch flow, advanced options moved into disclosures, single-row auth roles, and setup warnings (including a non-blocking no-auth warning) only appear from Check setup.
  • Desktop: cookie-file roles get a Browse file picker; About shows the VERDICT logo with runtime versions.
  • CLI gains a diff command for retest comparisons between runs.
  • Scope fix: explicit non-catch-all allow prefixes can carve exceptions over blocked paths, while the default catch-all allow no longer bypasses deny prefixes.
  • Runtime: Electron 44.3.0 -> 44.7.0; macOS installer workflow added (separate unsigned prerelease).
  • Release tooling: installer verification now correctly counts zero-byte payloads shipped upstream (hono/cn ESM stubs).

Upgrade

  • Close VERDICT and run the new Windows installer.
  • No saved-assessment or settings migration is required; existing runs and settings carry over.

Validation

Known limitations

  • macOS build is unsigned and published separately as a prerelease; it is not covered by this Windows release validation.
  • Live model checks exercised the OpenAI-compatible provider path (OpenCode Go, deepseek-v4.1-flash); the claude-cli subscription path is covered by unit tests only.
  • A full end-to-end paid-model assessment was not run against a live target; probe behavior is covered by local fixture tests.

v2026.10.1-macos — macOS arm64 build (unsigned, local)

Choose a tag to compare

@vvts-alpha vvts-alpha released this 19 Oct 03:46

Summary

macOS arm64 build of the VERDICT 2026.10.1 desktop app, packaged locally as a dmg.

This is not an output of the Windows release pipeline (promote-release.cjs) — it is a local electron-builder build uploaded as a convenience asset. The tagged source commit 7e002cc is functionally identical to d2f03b7 for shipped app code; the two later commits touch only a test fixture and a CI workflow.

Changes since v2026.9.12

  • WebUI migrated to Tailwind v4 + local shadcn/ui with a refreshed layout
  • Streamlined New Assessment launcher; per-row actions menu (Duplicate) on Projects
  • Cookie-file picker, About logo, warn-ready check display, bundled app icon
  • GraphQL operations captured as first-class inventory entries; expanded probe catalog (GraphQL ops, authz, ws, desync)
  • veritas diff CLI for retest comparisons; fingerprint updates; auth-isolated HTTP fixtures
  • Electron 44.7.0

Upgrade

Download VERDICT-2026.10.1-arm64.dmg, open it, and drag VERDICT to Applications.

The app is unsigned (no Developer ID certificate). On first launch either right-click → Open, or run:

xattr -cr /Applications/VERDICT.app

Validation

  • Built with Node 24 + pnpm 9.15.4 + electron-builder 26.15.3 on macOS (arm64), unchanged lockfile
  • Launches on macOS arm64: in-process server starts (server on http://127.0.0.1:*), main window renders the new WebUI (screenshot-verified)
  • Not run through the Windows CI gate suite (check-desktop-bundle); treat accordingly

Known limitations

  • Unsigned / adhoc-signed: Gatekeeper will warn until quarantine is removed (see Upgrade)
  • arm64 only — no x64 or universal build
  • Playwright automation browser is not bundled; point Settings at an installed Chrome/Edge (VERDICT_BROWSER_PATH)
  • Local build; the official release (when published via promote-release.cjs) is produced from the CI-tested Windows installer

v2026.9.12 — Preserve unfinished recon and prevent premature survey completion.

Choose a tag to compare

@vvts-alpha vvts-alpha released this 10 Sep 19:15

Summary

Preserve unfinished recon and prevent premature survey completion.

Changes

  • Survey completion now refuses unvisited in-scope URLs unless the configured survey cap has been reached.
  • Interrupted recon resumes its remaining work after mapping; completed recon is persisted and not repeated.
  • Windows builds use locked production dependencies, preserve hidden files during artifact transfer, verify downloaded installer contents against the tested app, and check packaged GUI, API authentication, CLI and native SDK startup.
  • Release promotion requires matching successful CI, Windows live checks, measured size budgets and verified upload hashes. README downloads now follow the latest release.

Upgrade

  • Close VERDICT and install the new Windows x64 installer.
  • Saved settings and assessments are retained. Resume an interrupted recon assessment to continue its remaining work.

Validation

Known limitations

  • Live model checks cover the configured OpenAI-compatible provider using a local no-op tool and no target assessment data. A full Claude-driven assessment was not run.
  • Windows x64 was validated. Linux/macOS installers and interactive Windows upgrade/uninstall were not exercised by this release automation.

v2026.9.11-1 — Windows packaging fix

Choose a tag to compare

@vvts-alpha vvts-alpha released this 10 Sep 18:07

Fixes oversized Windows installers by removing two Linux-only Claude SDK executables accidentally included in v2026.9.11. The Windows SDK executable and existing provider support are retained.

The packaging hook removes foreign SDK platforms and architectures from the Windows output and stops packaging if the required Windows executable is missing or invalid. The Windows builder version is pinned for reproducibility.

Validation: packaging regression tests pass; the Windows packaged app was checked for startup, settings IPC, local API authentication, CLI startup, native SDK startup, live OpenCodeGo connection and a local no-op tool roundtrip. Installer contents and checksums were inspected before publication.

v2026.9.11 — OpenCodeGo connectivity and security fixes

Choose a tag to compare

@vvts-alpha vvts-alpha released this 10 Sep 17:57

Fixes OpenCodeGo connections rejected with HTTP 400 despite a valid model and API key. VERDICT now sends an identifying user agent and a stable x-opencode-session across tool turns, retries and context summaries. Connection checks show safe, actionable failure categories, including missing Burp Audit authentication.

Security and reliability changes:

  • Update Electron to 44.3.0 and vulnerable dependencies; installer version follows the pinned runtime. Dependency audit reports no known vulnerabilities.
  • Authenticate desktop HTTP/WebSocket traffic with a per-launch credential and reject requests from other origins. Invalid URL encodings return request errors instead of terminating the server.
  • Pause incomplete model stages and keep unfinished screens queued. Preserve findings and full attack plans on resume, including legacy plan migration and coverage enforcement.
  • Respect private Public Suffix List boundaries when deriving scope; show allowed hosts before launching. Existing saved/explicit scopes are retained and should be reviewed before resuming older assessments.
  • Add model context limits and automatic compaction. Remove retired ASR commands and views; the API specification UI remains marked Coming Soon.

Validation: workspace build and typecheck passed; 721 automated tests passed; dependency audit found zero advisories. Tested Electron 44 on Linux and the Windows packaged application, including API authentication, IPC, CLI child startup, and live OpenCodeGo deepseek-v4-flash connection and tool-result roundtrip. Linux Electron checks also covered WebSockets, target-session isolation, PDF export and scope preview. Real model checks used a local no-op tool and sent no assessment data. Burp without an Audit token correctly reports HTTP 401/403.

VERDICT 2026.9.7 — Windows desktop

Choose a tag to compare

@vvts-alpha vvts-alpha released this 07 Sep 02:01

VERDICT 2026.9.7 updates the Windows desktop workflow, report review, and provider settings.

What's included

  • Model Settings now offers Claude / OpenCodeGo / OrcaRouter / Other. API presets prefill editable URLs; previous OpenAI-compatible settings migrate without losing the existing endpoint, key, or model IDs.
  • HTML and PDF reports use a save dialog, with the assessment ID in the suggested filename. Exporting keeps the assessment window open.
  • Final findings review consolidates supported duplicate SQLi, DOM XSS, address-write, and JWT observations while retaining the original records and evidence. Upload acceptance alone is not treated as executed Stored XSS; negative wallet deposits no longer imply a proven payout or theft.
  • Setup connection checks, API-spec assessment from OpenAPI/Swagger JSON, captured-session continuation, and corrected tested/excluded/unfinished counts are included since 2026.9.3.
  • Burp active scans submit one task at a time, wait for completion, and save results before the next submission. Errors, pauses, and timeouts stop further submissions.

Install and upgrade

Download VERDICT.Setup.2026.9.7.exe and run it on Windows x64. Close VERDICT before installing. Settings and assessments are stored under %APPDATA%\VERDICT.

The runtime is included; Node.js and pnpm are not required. Install Chrome or Edge for automated browsing and PDF export. Provider credentials or an authenticated Windows Claude CLI are still required. See WINDOWS-QUICKSTART.md and verify the installer against SHA256SUMS.txt.

For optional Burp Audit REST scanning, update/reload the attached verdict-burp-audit.jar (serial API v0.2.0). The app refuses the older extension for sequential submission.

Validation and limits

CI: workspace build, typecheck, and 753 tests passed. Windows build passed. Windows x64 runtime checks passed for SQLite, bundled CLI startup, renderer/API serving, provider mapping, and real HTML/PDF report output. Installer payload checks matched the tested app. Electron UI checks covered provider selection/persistence and report save/cancel behavior.

OpenCodeGo / omen-alpha was used for the operator's local Juice Shop assessment: 154 mapped screens marked scanned, with an operator-reported actual charge of US$9.92. The review notes distinguish revised report counts from independently validated unique vulnerabilities.

OrcaRouter's preset, persistence, and transport mapping were tested; a live OrcaRouter API assessment has not been run. Provider/model availability depends on your account. The installer remains unsigned and does not include automatic updates.

VERDICT 2026.9.3 — Windows desktop

Choose a tag to compare

@vvts-alpha vvts-alpha released this 05 Sep 08:49

VERDICT 2026.9.3 — Windows desktop

First GitHub Release of the VERDICT desktop app for Windows x64.

Download and start

Download VERDICT.Setup.2026.9.3.exe below. See WINDOWS-QUICKSTART.md for installation and first-run configuration; SHA256SUMS.txt contains the installer checksum.

The packaged app includes its runtime: Node.js and pnpm are not required. Install Chrome or Edge for automated assessments, and configure your LLM provider in Settings → Models. Claude CLI mode requires a separately installed, authenticated Windows claude command; an OpenAI-compatible endpoint can be configured in Settings instead.

Included

  • One desktop window for assessments, live progress, findings, evidence, and reports.
  • An embedded Browser tab for manual login and session capture, including cookies and localStorage.
  • Settings for LLM providers, Deep/Light models, proxy, browser, Burp, and out-of-band callbacks.
  • Reusable Operator context, clearer assessment creation, a Main navigation button, and version information under About.
  • The staged assessment engine with evidence confirmation and findings QA.

Build provenance

The installer is the unchanged artifact from Windows build 33653069591, built from commit f3e0aff0315f08bd765fa30796356a2b79ce728a. The build and test CI for that commit passed.

Additional verification:

  • Windows application launch confirmed by the operator.
  • The installer's extracted Windows x64 runtime passed SQLite write/read, bundled CLI child-process startup, desktop HTML/JavaScript serving, and the assessments API checks (Electron 38.0.0 / Node 22.18.0).
  • A complete live assessment on this release was not part of these packaging checks.

Known limitations

  • The installer is unsigned; Windows may show an unknown-publisher or SmartScreen warning.
  • Automatic updates, a first-run setup wizard, and a bundled automation browser are not included.
  • After an assessment process stops at an authentication wall, capture the session and resume from Main.
  • Use only targets you are authorized to test. Provider accounts and optional Burp services are configured separately.