Command Palette

Navigate
Projects
Writing
Actions
↑↓Navigate↵Select
EscClose
Work / All work2026 // Security / DevTools

Veylqorix — Code-security gate platform

new

Veylqorix is a code-security and quality gate platform built as a TypeScript/pnpm monorepo plus a standalone Rust security engine. It orchestrates real scanners — dependency auditing, secret detection, license checks, static analysis, and test suites — normalizes their output into one finding schema, and emits a deterministic gate decision (PASS / FAIL / INCOMPLETE, stable exit codes 0/1/2) with auditable evidence. CLI, MCP server, API, GitHub Actions, and agent skills are all thin adapters over that single engine, and an agent-safe policy means an AI agent can't suppress a finding. MIT-licensed, with sample data clearly labeled as sample.

TypeScriptRustpnpmMCPNode.jsGitHub ActionsAPI
Veylqorix system architecture diagram: the Rust core engine orchestrating security scanners (dependency audit, secret detection, static analysis, quality checks), the policy gate, findings persistence, reports, backend API, web dashboard, MCP server, and GitHub Actions integration
Veylqorix — Code-security gate platform2026

Live demo

The running product, embedded below. If it refuses to load in an embed, the live site opens in a new tab.

Problem

Security gates are either single scanners or brittle scripts: findings land in a dozen different formats, skipped or errored checks quietly count as clean, and teams can't prove what a gate actually checked.

Solution

Veylqorix is a TypeScript/pnpm monorepo plus a standalone Rust security engine. The engine runs real scanners — dependency auditing, secret detection, license checks, static analysis, test suites — normalizes every result into one finding schema, and produces a deterministic gate decision with stable exit codes (0 PASS, 1 FAIL, 2 INCOMPLETE). CLI, MCP server (6 tools), API, GitHub Actions and skills are all thin adapters over that single engine, and an agent-safe policy means an AI agent can't suppress a finding.

Decisions I made

  • 01INCOMPLETE as its own exit code (2) — a skipped or errored mandatory check degrades coverage and can never read as a clean pass; most pipelines collapse 'we didn't look' into green.
  • 02Rust engine, TS adapters — the verdict has to be identical whether it's made in CI, by the CLI, or by an agent calling the MCP tools; one engine, six thin adapters.
  • 03Agent-safe policy — suppression and policy edits come from a human-owned channel only; an agent can request a review, never remove a finding, and the audit trail records every change.
  • 04Sample data labeled as sample — presenting a synthetic scan as live results would erode the exact trust the product is trying to build.

Outcome

No unsafe code gets through, and no silent gaps either — a skipped mandatory check or scanner error degrades coverage and is marked SKIPPED/ERROR, never mistaken for a clean pass. Sample data is explicitly labeled, never presented as live results.

This project is part of the work I ship as a developer-tools engineer — see the freelance services page for project types, process, and how to start.

“The soul becomes dyed with the color of its thoughts.”
Marcus Aurelius
Soumya Chakraborty status avatar

Away from the keyboard.

Drop a mail, I'll reply when I'm back.

Made with <3 and lots of coffee.