Software & Agentic Systems

This topic focuses on architecture, reliability, deterministic workflows, review systems, and the engineering required to make agentic software operationally trustworthy.

• ساختن یک LLM Judge کم‌نویز برای بازبینی Pull Requestها: معماری K2 · Building a High-Signal LLM Judge for GitHub Pull Requests: The Architecture Behind K2 — A practical architecture for secure, SHA-aware, multi-agent pull-request review with evidence gates, low-noise inline findings, and a separate remediation loop.

• طراحی اسکیل‌های بزرگ ایجنت به‌صورت سیستم‌های قطعی و فازمحور · Designing Large Agent Skills as Deterministic, Phase-Oriented Systems — A reference architecture for explicit phases, lazy loading, bounded side effects, resumable handoffs, and testable agent execution.

Some work also appears under Artificial Intelligence because the subjects overlap. Each article still has one canonical URL.