arabic-rtl-agent-security-lab
Twenty executable Arabic, RTL, Arabizi, and mixed-language cases for goal hijacking, tool misuse, memory poisoning, Unicode controls, and synthetic data boundaries.
Mirogate Open Source
Open-source projects that show how Mirogate thinks about reliable systems.
These projects are designed to be useful on their own: executable Arabic/RTL agent-security tests, secure AI engineering frameworks, cascaded secure-coding and performance skills, public trust audits, Worker abuse defense, browser risk signals, lockout-safe cloud setup, migration labs, and bilingual enterprise forms.
secure-ai-engineering-framework
A high-security AI engineering framework: GPT-5.5 for code, Claude Opus for review, local models for sensitive data, private context packs, synthetic mocks, and secure-code checks.
secure-code-skill-cascade
A cascaded, editable secure-coding skill system for AI agents: OWASP, CWE, auth, input, data, dependencies, cloud, and AI-agent security modules.
web-performance-skill-cascade
A cascaded web performance skill system and CLI toolkit for AI agents: Lighthouse analysis, Core Web Vitals, budgets, JavaScript, media, caching, and CI checks.
web-security-baseline-audit
A small CLI that checks whether public pages expose the trust basics search engines, AI crawlers, and security reviewers expect to see.
cloudflare-worker-abuse-defense-kit
Worker patterns for public forms and lightweight APIs: native Workers rate limiting, Turnstile verification, honeypots, freshness checks, and privacy-minimized audit events.
fingerprintjs-risk-signal-lab
An unofficial defensive lab for turning browser visitor and abuse signals into explainable server-side decisions.
secure-droplet-bootstrap
A lockout-safe hardening kit for new DigitalOcean Ubuntu droplets: SSH, firewall, updates, baseline checks, and a Codex skill for assisted setup.
fingerprintjs-v6-readiness-lab
An unofficial readiness checklist and contract-test lab for teams preparing browser-identification code for a future major FingerprintJS upgrade.
bilingual-enterprise-form-kit
Arabic and English form patterns with RTL/LTR behavior, validation, honeypot protection, rate-limit examples, and accessible labels.
Credibility standard
Each project needs to be boring in the right ways.
- Clear README, license, security notes, and changelog from day one.
- Small test or verification command so a clean clone can prove the project works.
- Article and social posts that explain tradeoffs without overstating capability.
- Maintenance signals: issue templates, release notes, and visible updates after launch.