yanchuk

What 0RK Can Learn from Claude Code's Architecture

Analysis by a senior architect who knows both systems. Source: Claude Code CLI agent source code architecture deep dive.

---

High-Value Opportunities

1. Dream Task → The Missing Consistency Monitor

yanchuk

Claude Code Agent — Complete Architecture Deep Dive

A comprehensive analysis of the Claude Code CLI agent source code. Covers every subsystem: core loop, memory, orchestration, permissions, UI, MCP, skills, and more. Includes ASCII diagrams, design patterns, and ideas for building agent orchestrator systems.

---

Table of Contents

yanchuk

Tformance LLM Prompts

Prompt version v9.0.0 | Schema version v6.0.0

Tformance is a SaaS platform helping CTOs see how their team is actually delivering -- and whether AI coding tools are improving that delivery. We use LLMs to analyze every pull request and generate weekly team insights.

This gist contains the full production prompts powering two LLM pipelines:

yanchuk

Wi-Fi Router Requirements — Kyiv, Stalinka Apartment

Context

Parents' apartment in Kyiv: old "stalinka" building with thick brick/concrete walls. 2 rooms + kitchen, fiber optic internet (500-1000 Mbit/s). Frequent power outages → router runs on backup power (UPS/powerbank), so low power consumption is critical.

Requirements

| Parameter | Value |

yanchuk

GitHub Copilot Code Review Instructions — Zealth

You are a senior reviewer for a financial planning SaaS platform. Users trust this app with their financial data and projections. Your job is to catch what automation can't: architecture violations, correctness bugs, and security risks.

Only comment when you have HIGH CONFIDENCE (>80%) the issue is real. No hedging, no "you might want to consider." If you're not sure, don't comment. False positives create noise and erode trust. No one likes a reply guy.

Maximum 7 comments per review. Prioritize: security > correctness > architecture > performance. Never comment on unchanged code or suggest unrelated refactors.

---

yanchuk

Ski Equipment Price Research - Polish & EU Market, Spring 2026

Research Date: 2026-02-21 Items Researched: 60 (6 categories x 10 items) + 15 alternative finds Items Found in Poland: 49/60 Items Not Available in Poland: 7 (Trew Gear, Obermeyer, Glade, Flylow, Kinco, Pret Cynic X2, Helikon-Tex Merino Balaclava doesn't exist) Items with Limited PL Availability: 4 (Anon Prime MIPS, Gordini GTX, Goodr Snow G, Blackyak Gurla)

---

yanchuk

Overview

This PR adds a React Native (Expo) mobile app to the monorepo that wraps the Zealth web portal inside a native shell. The portal renders in a WebView; the native shell adds features that browsers cannot provide: biometric lock, native screen transitions, deep linking, and App Store presence.

Reading time: ~35 minutes. This description is written for experienced web engineers who haven't worked with React Native before. It covers: how the architecture works, what's different from web development, how Face ID and routing work under the hood, how to run and test the app, what's needed for App Store approval, Android readiness, chart library options, full native migration feasibility, and why we chose Expo.

---

Table of Contents

yanchuk

How Tformance Detects AI-Assisted Code (AI Gen %)

Approach: Disclosure-Based Detection (3 Layers)

We use a multi-signal detection system that analyzes PR metadata, not code diffs. This is intentional — stylometric code analysis has proven unreliable across the industry.

Layer 1: Pattern Matching (100+ regex rules)

• Scans PR titles, descriptions, and commit messages

yanchuk

Firestore Schema Migration Strategy

Overview

Firestore is "schemaless" - it doesn't enforce document structure. As our app evolves, we need a strategy for:

• Adding/removing fields from existing documents
• Renaming or restructuring data
• Ensuring consistency across collections

This document compares available approaches and provides implementation guidance.

yanchuk

# Identity

You are an engineering metrics analyst helping CTOs understand their team's performance and make data-driven decisions.

Your goal is to surface the single most important insight from the metrics data and provide an actionable recommendation.

# Instructions

## Task